diff --git a/art/estimators/classification/deep_partition_ensemble.py b/art/estimators/classification/deep_partition_ensemble.py index 09b943589f..a1779083b4 100644 --- a/art/estimators/classification/deep_partition_ensemble.py +++ b/art/estimators/classification/deep_partition_ensemble.py @@ -95,10 +95,12 @@ def __init__( ) if hash_function is None: - def hash_function(x): return int(np.sum(x)) % ensemble_size + if not isinstance(hash_function, Callable): + raise ValueError("hash_function must be callable") + self.hash_function = hash_function self.ensemble_size = ensemble_size diff --git a/notebooks/poisoning_defense_deep_partition_aggregation.ipynb b/notebooks/poisoning_defense_deep_partition_aggregation.ipynb index 442a8b1777..acb4c6ee89 100644 --- a/notebooks/poisoning_defense_deep_partition_aggregation.ipynb +++ b/notebooks/poisoning_defense_deep_partition_aggregation.ipynb @@ -128,7 +128,7 @@ { "data": { "text/plain": [ - "" + "" ] }, "execution_count": 4, @@ -176,7 +176,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "f57360b036904c72abfd151ed9507a3f", + "model_id": "3e50398472be4255b52c267d1461ea2c", "version_major": 2, "version_minor": 0 }, @@ -190,7 +190,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "814fa8e0a11542468f922908d5b41baa", + "model_id": "a860befc4231453f9a788160d4f6d18f", "version_major": 2, "version_minor": 0 }, @@ -219,7 +219,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "7fe361f7739445fa869afd3fcdc977fa", + "model_id": "df25857016e740f1816c54b5bf5507c2", "version_major": 2, "version_minor": 0 }, @@ -247,7 +247,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "c3b15a8e628c4bb3bdd8c1e234ae82e8", + "model_id": "00abadf513c340538a0150378302e7fb", "version_major": 2, "version_minor": 0 }, @@ -275,7 +275,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "6a22de3c51a14b0e81f4be6a5e480082", + "model_id": "8f90d38b4f8b4371a009173867ea7d2e", "version_major": 2, "version_minor": 0 }, @@ -303,7 +303,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "8664d522eb2b4bb0bb1be5c080986d5d", + "model_id": "bc9b20b9622641dd90fd72d3269fc56c", "version_major": 2, "version_minor": 0 }, @@ -331,7 +331,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "e916423fed1b4f568f29fcad83bfd9b6", + "model_id": "d5680eac57a74592b65845ee8336570b", "version_major": 2, "version_minor": 0 }, @@ -359,7 +359,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "308c9b99217b41e4b12d60e259813f35", + "model_id": "c1765d499b7b4f8ab37a5e0be62ca0e6", "version_major": 2, "version_minor": 0 }, @@ -387,7 +387,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "4cdb41f793cc43d7b52cf445aceebfc9", + "model_id": "8292f5a4a0ce4abf8b7a3eaedab663cd", "version_major": 2, "version_minor": 0 }, @@ -415,7 +415,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "9d5b2808bf1b49b081d1ada37d5d7979", + "model_id": "131b45f9cbd44661b045da9865ed58ac", "version_major": 2, "version_minor": 0 }, @@ -443,7 +443,7 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "a2565558de894641bd0698738be37485", + "model_id": "40ac0300c6004c10ae503e9af5c00dea", "version_major": 2, "version_minor": 0 }, @@ -471,7 +471,35 @@ { "data": { "application/vnd.jupyter.widget-view+json": { - "model_id": "9afab44d61de4041a8d5ba49166c94c7", + "model_id": "95517f5707d94b47b1f798c698624cf1", + "version_major": 2, + "version_minor": 0 + }, + "text/plain": [ + "PGD - Random Initializations: 0%| | 0/1 [00:00" ] @@ -533,7 +561,7 @@ "name": "stdout", "output_type": "stream", "text": [ - "Index: 5 Label: 9\n" + "Index: 290 Label: 9\n" ] } ], @@ -591,164 +619,164 @@ "text": [ "Train on 10000 samples\n", "Epoch 1/10\n", - "10000/10000 [==============================] - 1s 105us/sample - loss: 0.6879 - accuracy: 0.7870\n", + "10000/10000 [==============================] - 1s 109us/sample - loss: 0.6668 - accuracy: 0.7935\n", "Epoch 2/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.2275 - accuracy: 0.9339\n", + "10000/10000 [==============================] - 1s 53us/sample - loss: 0.2129 - accuracy: 0.9366\n", "Epoch 3/10\n", - "10000/10000 [==============================] - 1s 52us/sample - loss: 0.1439 - accuracy: 0.9557\n", + "10000/10000 [==============================] - 1s 54us/sample - loss: 0.1514 - accuracy: 0.9544\n", "Epoch 4/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.1099 - accuracy: 0.9675\n", + "10000/10000 [==============================] - 1s 53us/sample - loss: 0.1092 - accuracy: 0.9691\n", "Epoch 5/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.0875 - accuracy: 0.9743\n", + "10000/10000 [==============================] - 1s 52us/sample - loss: 0.0860 - accuracy: 0.9740\n", "Epoch 6/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.0712 - accuracy: 0.9793\n", + "10000/10000 [==============================] - 1s 52us/sample - loss: 0.0668 - accuracy: 0.9794\n", "Epoch 7/10\n", - "10000/10000 [==============================] - 1s 52us/sample - loss: 0.0681 - accuracy: 0.9777\n", + "10000/10000 [==============================] - 1s 52us/sample - loss: 0.0636 - accuracy: 0.9789\n", "Epoch 8/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.0467 - accuracy: 0.9847\n", + "10000/10000 [==============================] - 1s 52us/sample - loss: 0.0554 - accuracy: 0.9822\n", "Epoch 9/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.0424 - accuracy: 0.9876\n", + "10000/10000 [==============================] - 1s 54us/sample - loss: 0.0441 - accuracy: 0.9871\n", "Epoch 10/10\n", - "10000/10000 [==============================] - 1s 51us/sample - loss: 0.0382 - accuracy: 0.9875\n", - "Train on 961 samples\n", + "10000/10000 [==============================] - 1s 53us/sample - loss: 0.0366 - accuracy: 0.9878\n", + "Train on 963 samples\n", "Epoch 1/10\n", - "961/961 [==============================] - 9s 9ms/sample - loss: 1.8653 - accuracy: 0.4339\n", + "963/963 [==============================] - 9s 10ms/sample - loss: 1.9348 - accuracy: 0.3593\n", "Epoch 2/10\n", - "961/961 [==============================] - 0s 55us/sample - loss: 0.9416 - accuracy: 0.6785\n", + "963/963 [==============================] - 0s 55us/sample - loss: 1.0212 - accuracy: 0.6646\n", "Epoch 3/10\n", - "961/961 [==============================] - 0s 54us/sample - loss: 0.6542 - accuracy: 0.7908\n", + "963/963 [==============================] - 0s 54us/sample - loss: 0.6831 - accuracy: 0.7850\n", "Epoch 4/10\n", - "961/961 [==============================] - 0s 52us/sample - loss: 0.4830 - accuracy: 0.8408\n", + "963/963 [==============================] - 0s 58us/sample - loss: 0.5596 - accuracy: 0.8245\n", "Epoch 5/10\n", - "961/961 [==============================] - 0s 53us/sample - loss: 0.3970 - accuracy: 0.8730\n", + "963/963 [==============================] - 0s 53us/sample - loss: 0.4262 - accuracy: 0.8640\n", "Epoch 6/10\n", - "961/961 [==============================] - 0s 53us/sample - loss: 0.3229 - accuracy: 0.9032\n", + "963/963 [==============================] - 0s 53us/sample - loss: 0.3805 - accuracy: 0.8795\n", "Epoch 7/10\n", - "961/961 [==============================] - 0s 53us/sample - loss: 0.2604 - accuracy: 0.9126\n", + "963/963 [==============================] - 0s 56us/sample - loss: 0.3167 - accuracy: 0.9034\n", "Epoch 8/10\n", - "961/961 [==============================] - 0s 53us/sample - loss: 0.2045 - accuracy: 0.9303\n", + "963/963 [==============================] - 0s 54us/sample - loss: 0.2656 - accuracy: 0.9211\n", "Epoch 9/10\n", - "961/961 [==============================] - 0s 53us/sample - loss: 0.1648 - accuracy: 0.9542\n", + "963/963 [==============================] - 0s 53us/sample - loss: 0.2042 - accuracy: 0.9398\n", "Epoch 10/10\n", - "961/961 [==============================] - 0s 54us/sample - loss: 0.1564 - accuracy: 0.9511\n", - "Train on 1040 samples\n", + "963/963 [==============================] - 0s 53us/sample - loss: 0.1519 - accuracy: 0.9470\n", + "Train on 1000 samples\n", "Epoch 1/10\n", - "1040/1040 [==============================] - 1s 572us/sample - loss: 1.8364 - accuracy: 0.3808\n", + "1000/1000 [==============================] - 9s 9ms/sample - loss: 1.8922 - accuracy: 0.3860\n", "Epoch 2/10\n", - "1040/1040 [==============================] - 0s 54us/sample - loss: 0.9395 - accuracy: 0.7038\n", + "1000/1000 [==============================] - 0s 55us/sample - loss: 1.0128 - accuracy: 0.6770\n", "Epoch 3/10\n", - "1040/1040 [==============================] - 0s 53us/sample - loss: 0.6502 - accuracy: 0.7913\n", + "1000/1000 [==============================] - 0s 52us/sample - loss: 0.6467 - accuracy: 0.7990\n", "Epoch 4/10\n", - "1040/1040 [==============================] - 0s 53us/sample - loss: 0.5287 - accuracy: 0.8279\n", + "1000/1000 [==============================] - 0s 53us/sample - loss: 0.4744 - accuracy: 0.8530\n", "Epoch 5/10\n", - "1040/1040 [==============================] - 0s 54us/sample - loss: 0.4402 - accuracy: 0.8587\n", + "1000/1000 [==============================] - 0s 67us/sample - loss: 0.3595 - accuracy: 0.8960\n", "Epoch 6/10\n", - "1040/1040 [==============================] - 0s 53us/sample - loss: 0.3502 - accuracy: 0.8913\n", + "1000/1000 [==============================] - 0s 53us/sample - loss: 0.3195 - accuracy: 0.9000\n", "Epoch 7/10\n", - "1040/1040 [==============================] - 0s 55us/sample - loss: 0.3019 - accuracy: 0.9096\n", + "1000/1000 [==============================] - 0s 53us/sample - loss: 0.2663 - accuracy: 0.9220\n", "Epoch 8/10\n", - "1040/1040 [==============================] - 0s 54us/sample - loss: 0.2472 - accuracy: 0.9163\n", + "1000/1000 [==============================] - 0s 53us/sample - loss: 0.2528 - accuracy: 0.9260\n", "Epoch 9/10\n", - "1040/1040 [==============================] - 0s 53us/sample - loss: 0.2027 - accuracy: 0.9394\n", + "1000/1000 [==============================] - 0s 52us/sample - loss: 0.2115 - accuracy: 0.9390\n", "Epoch 10/10\n", - "1040/1040 [==============================] - 0s 54us/sample - loss: 0.1679 - accuracy: 0.9510\n", - "Train on 988 samples\n", + "1000/1000 [==============================] - 0s 54us/sample - loss: 0.1582 - accuracy: 0.9570\n", + "Train on 1057 samples\n", "Epoch 1/10\n", - "988/988 [==============================] - 9s 9ms/sample - loss: 1.8526 - accuracy: 0.3978\n", + "1057/1057 [==============================] - 9s 9ms/sample - loss: 1.8447 - accuracy: 0.3860\n", "Epoch 2/10\n", - "988/988 [==============================] - 0s 54us/sample - loss: 0.9549 - accuracy: 0.6812\n", + "1057/1057 [==============================] - 0s 59us/sample - loss: 0.9678 - accuracy: 0.6717\n", "Epoch 3/10\n", - "988/988 [==============================] - 0s 53us/sample - loss: 0.6674 - accuracy: 0.7864\n", + "1057/1057 [==============================] - 0s 56us/sample - loss: 0.7032 - accuracy: 0.7588\n", "Epoch 4/10\n", - "988/988 [==============================] - 0s 53us/sample - loss: 0.5295 - accuracy: 0.8219\n", + "1057/1057 [==============================] - 0s 56us/sample - loss: 0.5346 - accuracy: 0.8269\n", "Epoch 5/10\n", - "988/988 [==============================] - 0s 53us/sample - loss: 0.4049 - accuracy: 0.8715\n", + "1057/1057 [==============================] - 0s 56us/sample - loss: 0.4131 - accuracy: 0.8675\n", "Epoch 6/10\n", - "988/988 [==============================] - 0s 52us/sample - loss: 0.2809 - accuracy: 0.9028\n", + "1057/1057 [==============================] - 0s 55us/sample - loss: 0.3383 - accuracy: 0.8931\n", "Epoch 7/10\n", - "988/988 [==============================] - 0s 53us/sample - loss: 0.2446 - accuracy: 0.9150\n", + "1057/1057 [==============================] - 0s 55us/sample - loss: 0.2952 - accuracy: 0.9139\n", "Epoch 8/10\n", - "988/988 [==============================] - 0s 52us/sample - loss: 0.2142 - accuracy: 0.9231\n", + "1057/1057 [==============================] - 0s 57us/sample - loss: 0.2469 - accuracy: 0.9205\n", "Epoch 9/10\n", - "988/988 [==============================] - 0s 53us/sample - loss: 0.1767 - accuracy: 0.9393\n", + "1057/1057 [==============================] - 0s 55us/sample - loss: 0.2121 - accuracy: 0.9338\n", "Epoch 10/10\n", - "988/988 [==============================] - 0s 51us/sample - loss: 0.1370 - accuracy: 0.9514\n", - "Train on 983 samples\n", + "1057/1057 [==============================] - 0s 61us/sample - loss: 0.1670 - accuracy: 0.9489\n", + "Train on 1032 samples\n", "Epoch 1/10\n", - "983/983 [==============================] - 9s 9ms/sample - loss: 1.8752 - accuracy: 0.3957\n", + "1032/1032 [==============================] - 9s 9ms/sample - loss: 1.8166 - accuracy: 0.4254\n", "Epoch 2/10\n", - "983/983 [==============================] - 0s 54us/sample - loss: 0.9625 - accuracy: 0.6948\n", + "1032/1032 [==============================] - 0s 60us/sample - loss: 0.9565 - accuracy: 0.6841\n", "Epoch 3/10\n", - "983/983 [==============================] - 0s 52us/sample - loss: 0.6611 - accuracy: 0.7792\n", + "1032/1032 [==============================] - 0s 57us/sample - loss: 0.6313 - accuracy: 0.7994\n", "Epoch 4/10\n", - "983/983 [==============================] - 0s 53us/sample - loss: 0.5149 - accuracy: 0.8311\n", + "1032/1032 [==============================] - 0s 58us/sample - loss: 0.4817 - accuracy: 0.8440\n", "Epoch 5/10\n", - "983/983 [==============================] - 0s 53us/sample - loss: 0.3772 - accuracy: 0.8789\n", + "1032/1032 [==============================] - 0s 58us/sample - loss: 0.4146 - accuracy: 0.8750\n", "Epoch 6/10\n", - "983/983 [==============================] - 0s 52us/sample - loss: 0.3482 - accuracy: 0.8901\n", + "1032/1032 [==============================] - 0s 56us/sample - loss: 0.3367 - accuracy: 0.8905\n", "Epoch 7/10\n", - "983/983 [==============================] - 0s 52us/sample - loss: 0.2651 - accuracy: 0.9196\n", + "1032/1032 [==============================] - 0s 59us/sample - loss: 0.2960 - accuracy: 0.9138\n", "Epoch 8/10\n", - "983/983 [==============================] - 0s 52us/sample - loss: 0.2234 - accuracy: 0.9318\n", + "1032/1032 [==============================] - 0s 57us/sample - loss: 0.2549 - accuracy: 0.9225\n", "Epoch 9/10\n", - "983/983 [==============================] - 0s 51us/sample - loss: 0.1985 - accuracy: 0.9349\n", + "1032/1032 [==============================] - 0s 55us/sample - loss: 0.2414 - accuracy: 0.9264\n", "Epoch 10/10\n", - "983/983 [==============================] - 0s 53us/sample - loss: 0.1622 - accuracy: 0.9471\n", - "Train on 1009 samples\n", + "1032/1032 [==============================] - 0s 56us/sample - loss: 0.1950 - accuracy: 0.9486\n", + "Train on 1020 samples\n", "Epoch 1/10\n", - "1009/1009 [==============================] - 9s 9ms/sample - loss: 1.8974 - accuracy: 0.4054\n", + "1020/1020 [==============================] - 9s 9ms/sample - loss: 1.8730 - accuracy: 0.4039\n", "Epoch 2/10\n", - "1009/1009 [==============================] - 0s 53us/sample - loss: 0.9712 - accuracy: 0.6868\n", + "1020/1020 [==============================] - 0s 56us/sample - loss: 0.9344 - accuracy: 0.7029\n", "Epoch 3/10\n", - "1009/1009 [==============================] - 0s 56us/sample - loss: 0.6651 - accuracy: 0.7919\n", + "1020/1020 [==============================] - 0s 54us/sample - loss: 0.6400 - accuracy: 0.7912\n", "Epoch 4/10\n", - "1009/1009 [==============================] - 0s 64us/sample - loss: 0.4878 - accuracy: 0.8404\n", + "1020/1020 [==============================] - 0s 53us/sample - loss: 0.5197 - accuracy: 0.8324\n", "Epoch 5/10\n", - "1009/1009 [==============================] - 0s 51us/sample - loss: 0.3912 - accuracy: 0.8840\n", + "1020/1020 [==============================] - 0s 52us/sample - loss: 0.4154 - accuracy: 0.8696\n", "Epoch 6/10\n", - "1009/1009 [==============================] - 0s 52us/sample - loss: 0.3520 - accuracy: 0.8870\n", + "1020/1020 [==============================] - 0s 54us/sample - loss: 0.3371 - accuracy: 0.8824\n", "Epoch 7/10\n", - "1009/1009 [==============================] - 0s 53us/sample - loss: 0.2473 - accuracy: 0.9207\n", + "1020/1020 [==============================] - 0s 55us/sample - loss: 0.2922 - accuracy: 0.9098\n", "Epoch 8/10\n", - "1009/1009 [==============================] - 0s 51us/sample - loss: 0.2203 - accuracy: 0.9316\n", + "1020/1020 [==============================] - 0s 62us/sample - loss: 0.2364 - accuracy: 0.9294\n", "Epoch 9/10\n", - "1009/1009 [==============================] - 0s 51us/sample - loss: 0.1791 - accuracy: 0.9415\n", + "1020/1020 [==============================] - 0s 52us/sample - loss: 0.1839 - accuracy: 0.9461\n", "Epoch 10/10\n", - "1009/1009 [==============================] - 0s 52us/sample - loss: 0.1321 - accuracy: 0.9604\n", - "Train on 1027 samples\n", + "1020/1020 [==============================] - 0s 51us/sample - loss: 0.1587 - accuracy: 0.9412\n", + "Train on 1005 samples\n", "Epoch 1/10\n", - "1027/1027 [==============================] - 9s 9ms/sample - loss: 1.8899 - accuracy: 0.3973\n", + "1005/1005 [==============================] - 9s 9ms/sample - loss: 1.8552 - accuracy: 0.4269\n", "Epoch 2/10\n", - "1027/1027 [==============================] - 0s 58us/sample - loss: 1.1001 - accuracy: 0.6368\n", + "1005/1005 [==============================] - 0s 55us/sample - loss: 1.0098 - accuracy: 0.6726\n", "Epoch 3/10\n", - "1027/1027 [==============================] - 0s 57us/sample - loss: 0.7573 - accuracy: 0.7507\n", + "1005/1005 [==============================] - 0s 57us/sample - loss: 0.6585 - accuracy: 0.7791\n", "Epoch 4/10\n", - "1027/1027 [==============================] - 0s 57us/sample - loss: 0.5729 - accuracy: 0.8208\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.4911 - accuracy: 0.8388\n", "Epoch 5/10\n", - "1027/1027 [==============================] - 0s 57us/sample - loss: 0.4595 - accuracy: 0.8530\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.3979 - accuracy: 0.8697\n", "Epoch 6/10\n", - "1027/1027 [==============================] - 0s 55us/sample - loss: 0.4034 - accuracy: 0.8608\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.2860 - accuracy: 0.9045\n", "Epoch 7/10\n", - "1027/1027 [==============================] - 0s 55us/sample - loss: 0.3359 - accuracy: 0.8968\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.2859 - accuracy: 0.9075\n", "Epoch 8/10\n", - "1027/1027 [==============================] - 0s 57us/sample - loss: 0.2833 - accuracy: 0.9026\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.2305 - accuracy: 0.9274\n", "Epoch 9/10\n", - "1027/1027 [==============================] - 0s 55us/sample - loss: 0.2264 - accuracy: 0.9172\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.1840 - accuracy: 0.9403\n", "Epoch 10/10\n", - "1027/1027 [==============================] - 0s 54us/sample - loss: 0.2377 - accuracy: 0.9221\n", - "Train on 975 samples\n", + "1005/1005 [==============================] - 0s 53us/sample - loss: 0.1750 - accuracy: 0.9433\n", + "Train on 968 samples\n", "Epoch 1/10\n", - "975/975 [==============================] - 9s 9ms/sample - loss: 1.8936 - accuracy: 0.4010\n", + "968/968 [==============================] - 9s 10ms/sample - loss: 1.8647 - accuracy: 0.3729\n", "Epoch 2/10\n", - "975/975 [==============================] - 0s 56us/sample - loss: 0.9852 - accuracy: 0.6872\n", + "968/968 [==============================] - 0s 58us/sample - loss: 0.9051 - accuracy: 0.7014\n", "Epoch 3/10\n", - "975/975 [==============================] - 0s 53us/sample - loss: 0.6404 - accuracy: 0.8031\n", + "968/968 [==============================] - 0s 55us/sample - loss: 0.6047 - accuracy: 0.7955\n", "Epoch 4/10\n", - "975/975 [==============================] - 0s 55us/sample - loss: 0.4868 - accuracy: 0.8308\n", + "968/968 [==============================] - 0s 55us/sample - loss: 0.4757 - accuracy: 0.8554\n", "Epoch 5/10\n", - "975/975 [==============================] - 0s 54us/sample - loss: 0.3950 - accuracy: 0.8779\n", + "968/968 [==============================] - 0s 55us/sample - loss: 0.3595 - accuracy: 0.8781\n", "Epoch 6/10\n", - "975/975 [==============================] - 0s 52us/sample - loss: 0.3274 - accuracy: 0.8964\n", + "968/968 [==============================] - 0s 65us/sample - loss: 0.3046 - accuracy: 0.9029\n", "Epoch 7/10\n" ] }, @@ -756,167 +784,167 @@ "name": "stdout", "output_type": "stream", "text": [ - "975/975 [==============================] - 0s 54us/sample - loss: 0.2622 - accuracy: 0.9159\n", + "968/968 [==============================] - 0s 55us/sample - loss: 0.2729 - accuracy: 0.9070\n", "Epoch 8/10\n", - "975/975 [==============================] - 0s 54us/sample - loss: 0.2129 - accuracy: 0.9333\n", + "968/968 [==============================] - 0s 56us/sample - loss: 0.2170 - accuracy: 0.9277\n", "Epoch 9/10\n", - "975/975 [==============================] - 0s 54us/sample - loss: 0.1677 - accuracy: 0.9487\n", + "968/968 [==============================] - 0s 60us/sample - loss: 0.1661 - accuracy: 0.9473\n", "Epoch 10/10\n", - "975/975 [==============================] - 0s 54us/sample - loss: 0.1530 - accuracy: 0.9579\n", - "Train on 1028 samples\n", + "968/968 [==============================] - 0s 55us/sample - loss: 0.1444 - accuracy: 0.9473\n", + "Train on 986 samples\n", "Epoch 1/10\n", - "1028/1028 [==============================] - 9s 9ms/sample - loss: 1.8419 - accuracy: 0.4037\n", + "986/986 [==============================] - 9s 10ms/sample - loss: 1.8995 - accuracy: 0.3793\n", "Epoch 2/10\n", - "1028/1028 [==============================] - 0s 58us/sample - loss: 0.9608 - accuracy: 0.6839\n", + "986/986 [==============================] - 0s 57us/sample - loss: 0.9369 - accuracy: 0.6876\n", "Epoch 3/10\n", - "1028/1028 [==============================] - 0s 57us/sample - loss: 0.7111 - accuracy: 0.7626\n", + "986/986 [==============================] - 0s 60us/sample - loss: 0.6742 - accuracy: 0.7769\n", "Epoch 4/10\n", - "1028/1028 [==============================] - 0s 57us/sample - loss: 0.5721 - accuracy: 0.8200\n", + "986/986 [==============================] - 0s 54us/sample - loss: 0.5473 - accuracy: 0.8479\n", "Epoch 5/10\n", - "1028/1028 [==============================] - 0s 57us/sample - loss: 0.4415 - accuracy: 0.8512\n", + "986/986 [==============================] - 0s 53us/sample - loss: 0.4056 - accuracy: 0.8763\n", "Epoch 6/10\n", - "1028/1028 [==============================] - 0s 59us/sample - loss: 0.4464 - accuracy: 0.8551\n", + "986/986 [==============================] - 0s 57us/sample - loss: 0.3646 - accuracy: 0.8824\n", "Epoch 7/10\n", - "1028/1028 [==============================] - 0s 55us/sample - loss: 0.3433 - accuracy: 0.8920\n", + "986/986 [==============================] - 0s 54us/sample - loss: 0.3166 - accuracy: 0.8976\n", "Epoch 8/10\n", - "1028/1028 [==============================] - 0s 60us/sample - loss: 0.2768 - accuracy: 0.9095\n", + "986/986 [==============================] - 0s 54us/sample - loss: 0.2728 - accuracy: 0.9128\n", "Epoch 9/10\n", - "1028/1028 [==============================] - 0s 69us/sample - loss: 0.2491 - accuracy: 0.9193\n", + "986/986 [==============================] - 0s 53us/sample - loss: 0.2344 - accuracy: 0.9290\n", "Epoch 10/10\n", - "1028/1028 [==============================] - 0s 55us/sample - loss: 0.2224 - accuracy: 0.9280\n", - "Train on 1013 samples\n", + "986/986 [==============================] - 0s 53us/sample - loss: 0.1930 - accuracy: 0.9351\n", + "Train on 991 samples\n", "Epoch 1/10\n", - "1013/1013 [==============================] - 9s 9ms/sample - loss: 1.8913 - accuracy: 0.3889\n", + "991/991 [==============================] - 10s 10ms/sample - loss: 1.9182 - accuracy: 0.3824\n", "Epoch 2/10\n", - "1013/1013 [==============================] - 0s 55us/sample - loss: 0.9575 - accuracy: 0.6960\n", + "991/991 [==============================] - 0s 55us/sample - loss: 1.0203 - accuracy: 0.6842\n", "Epoch 3/10\n", - "1013/1013 [==============================] - 0s 52us/sample - loss: 0.6871 - accuracy: 0.7996\n", + "991/991 [==============================] - 0s 53us/sample - loss: 0.7130 - accuracy: 0.7709\n", "Epoch 4/10\n", - "1013/1013 [==============================] - 0s 52us/sample - loss: 0.5204 - accuracy: 0.8411\n", + "991/991 [==============================] - 0s 53us/sample - loss: 0.5373 - accuracy: 0.8385\n", "Epoch 5/10\n", - "1013/1013 [==============================] - 0s 52us/sample - loss: 0.4193 - accuracy: 0.8786\n", + "991/991 [==============================] - 0s 52us/sample - loss: 0.4151 - accuracy: 0.8840\n", "Epoch 6/10\n", - "1013/1013 [==============================] - 0s 53us/sample - loss: 0.3548 - accuracy: 0.8954\n", + "991/991 [==============================] - 0s 53us/sample - loss: 0.3558 - accuracy: 0.8799\n", "Epoch 7/10\n", - "1013/1013 [==============================] - 0s 53us/sample - loss: 0.2864 - accuracy: 0.9191\n", + "991/991 [==============================] - 0s 63us/sample - loss: 0.3103 - accuracy: 0.9082\n", "Epoch 8/10\n", - "1013/1013 [==============================] - 0s 53us/sample - loss: 0.2319 - accuracy: 0.9269\n", + "991/991 [==============================] - 0s 55us/sample - loss: 0.2769 - accuracy: 0.9092\n", "Epoch 9/10\n", - "1013/1013 [==============================] - 0s 53us/sample - loss: 0.1835 - accuracy: 0.9408\n", + "991/991 [==============================] - 0s 54us/sample - loss: 0.2179 - accuracy: 0.9314\n", "Epoch 10/10\n", - "1013/1013 [==============================] - 0s 51us/sample - loss: 0.1563 - accuracy: 0.9536\n", - "Train on 976 samples\n", + "991/991 [==============================] - 0s 54us/sample - loss: 0.1640 - accuracy: 0.9516\n", + "Train on 978 samples\n", "Epoch 1/10\n", - "976/976 [==============================] - 9s 10ms/sample - loss: 1.9272 - accuracy: 0.3750\n", + "978/978 [==============================] - 10s 10ms/sample - loss: 1.8435 - accuracy: 0.3875\n", "Epoch 2/10\n", - "976/976 [==============================] - 0s 59us/sample - loss: 1.0113 - accuracy: 0.6588\n", + "978/978 [==============================] - 0s 57us/sample - loss: 0.9792 - accuracy: 0.6748\n", "Epoch 3/10\n", - "976/976 [==============================] - 0s 53us/sample - loss: 0.6702 - accuracy: 0.7756\n", + "978/978 [==============================] - 0s 55us/sample - loss: 0.6562 - accuracy: 0.7924\n", "Epoch 4/10\n", - "976/976 [==============================] - 0s 69us/sample - loss: 0.5410 - accuracy: 0.8258\n", + "978/978 [==============================] - 0s 55us/sample - loss: 0.5470 - accuracy: 0.8395\n", "Epoch 5/10\n", - "976/976 [==============================] - 0s 54us/sample - loss: 0.4228 - accuracy: 0.8668\n", + "978/978 [==============================] - 0s 56us/sample - loss: 0.4449 - accuracy: 0.8661\n", "Epoch 6/10\n", - "976/976 [==============================] - 0s 54us/sample - loss: 0.3236 - accuracy: 0.8873\n", + "978/978 [==============================] - 0s 53us/sample - loss: 0.3623 - accuracy: 0.8845\n", "Epoch 7/10\n", - "976/976 [==============================] - 0s 56us/sample - loss: 0.2605 - accuracy: 0.9191\n", + "978/978 [==============================] - 0s 54us/sample - loss: 0.3123 - accuracy: 0.8937\n", "Epoch 8/10\n", - "976/976 [==============================] - 0s 53us/sample - loss: 0.2352 - accuracy: 0.9221\n", + "978/978 [==============================] - 0s 63us/sample - loss: 0.2351 - accuracy: 0.9315\n", "Epoch 9/10\n", - "976/976 [==============================] - 0s 53us/sample - loss: 0.1963 - accuracy: 0.9283\n", + "978/978 [==============================] - 0s 55us/sample - loss: 0.2165 - accuracy: 0.9356\n", "Epoch 10/10\n", - "976/976 [==============================] - 0s 53us/sample - loss: 0.1746 - accuracy: 0.9467\n", + "978/978 [==============================] - 0s 54us/sample - loss: 0.1690 - accuracy: 0.9581\n", "Train on 481 samples\n", "Epoch 1/10\n", - "481/481 [==============================] - 9s 19ms/sample - loss: 2.1913 - accuracy: 0.2287\n", + "481/481 [==============================] - 10s 20ms/sample - loss: 2.1775 - accuracy: 0.2183\n", "Epoch 2/10\n", - "481/481 [==============================] - 0s 59us/sample - loss: 1.5057 - accuracy: 0.6279\n", + "481/481 [==============================] - 0s 64us/sample - loss: 1.6444 - accuracy: 0.4990\n", "Epoch 3/10\n", - "481/481 [==============================] - 0s 54us/sample - loss: 0.9387 - accuracy: 0.7401\n", + "481/481 [==============================] - 0s 57us/sample - loss: 1.1730 - accuracy: 0.6133\n", "Epoch 4/10\n", - "481/481 [==============================] - 0s 54us/sample - loss: 0.7536 - accuracy: 0.7214\n", + "481/481 [==============================] - 0s 58us/sample - loss: 0.8438 - accuracy: 0.7256\n", "Epoch 5/10\n", - "481/481 [==============================] - 0s 55us/sample - loss: 0.5725 - accuracy: 0.8212\n", + "481/481 [==============================] - 0s 57us/sample - loss: 0.7197 - accuracy: 0.7630\n", "Epoch 6/10\n", - "481/481 [==============================] - 0s 54us/sample - loss: 0.4844 - accuracy: 0.8545\n", + "481/481 [==============================] - 0s 55us/sample - loss: 0.5893 - accuracy: 0.8212\n", "Epoch 7/10\n", - "481/481 [==============================] - 0s 54us/sample - loss: 0.4024 - accuracy: 0.8857\n", + "481/481 [==============================] - 0s 56us/sample - loss: 0.5090 - accuracy: 0.8337\n", "Epoch 8/10\n", - "481/481 [==============================] - 0s 56us/sample - loss: 0.3480 - accuracy: 0.8857\n", + "481/481 [==============================] - 0s 56us/sample - loss: 0.4340 - accuracy: 0.8565\n", "Epoch 9/10\n", - "481/481 [==============================] - 0s 55us/sample - loss: 0.3298 - accuracy: 0.8919\n", + "481/481 [==============================] - 0s 59us/sample - loss: 0.3860 - accuracy: 0.8690\n", "Epoch 10/10\n", - "481/481 [==============================] - 0s 73us/sample - loss: 0.2510 - accuracy: 0.9168\n", - "Train on 542 samples\n", + "481/481 [==============================] - 0s 56us/sample - loss: 0.3088 - accuracy: 0.9210\n", + "Train on 514 samples\n", "Epoch 1/10\n", - "542/542 [==============================] - 9s 17ms/sample - loss: 2.1298 - accuracy: 0.3081\n", + "514/514 [==============================] - 10s 19ms/sample - loss: 2.1181 - accuracy: 0.2704\n", "Epoch 2/10\n", - "542/542 [==============================] - 0s 61us/sample - loss: 1.3222 - accuracy: 0.6292\n", + "514/514 [==============================] - 0s 66us/sample - loss: 1.4391 - accuracy: 0.5564\n", "Epoch 3/10\n", - "542/542 [==============================] - 0s 57us/sample - loss: 0.9315 - accuracy: 0.6863\n", + "514/514 [==============================] - 0s 62us/sample - loss: 1.0817 - accuracy: 0.6712\n", "Epoch 4/10\n", - "542/542 [==============================] - 0s 56us/sample - loss: 0.7426 - accuracy: 0.7565\n", + "514/514 [==============================] - 0s 63us/sample - loss: 0.8721 - accuracy: 0.7121\n", "Epoch 5/10\n", - "542/542 [==============================] - 0s 59us/sample - loss: 0.5903 - accuracy: 0.8266\n", + "514/514 [==============================] - 0s 64us/sample - loss: 0.7518 - accuracy: 0.7626\n", "Epoch 6/10\n", - "542/542 [==============================] - 0s 56us/sample - loss: 0.4974 - accuracy: 0.8506\n", + "514/514 [==============================] - 0s 65us/sample - loss: 0.6227 - accuracy: 0.8054\n", "Epoch 7/10\n", - "542/542 [==============================] - 0s 62us/sample - loss: 0.3969 - accuracy: 0.8838\n", + "514/514 [==============================] - 0s 61us/sample - loss: 0.5887 - accuracy: 0.8191\n", "Epoch 8/10\n", - "542/542 [==============================] - 0s 58us/sample - loss: 0.3510 - accuracy: 0.8930\n", + "514/514 [==============================] - 0s 61us/sample - loss: 0.5054 - accuracy: 0.8658\n", "Epoch 9/10\n", - "542/542 [==============================] - 0s 63us/sample - loss: 0.2787 - accuracy: 0.9041\n", + "514/514 [==============================] - 0s 59us/sample - loss: 0.3586 - accuracy: 0.8988\n", "Epoch 10/10\n", - "542/542 [==============================] - 0s 59us/sample - loss: 0.2160 - accuracy: 0.9262\n", - "Train on 512 samples\n", + "514/514 [==============================] - 0s 60us/sample - loss: 0.3614 - accuracy: 0.8911\n", + "Train on 523 samples\n", "Epoch 1/10\n", - "512/512 [==============================] - 1s 1ms/sample - loss: 2.1561 - accuracy: 0.2656\n", + "523/523 [==============================] - 10s 18ms/sample - loss: 2.1203 - accuracy: 0.3327\n", "Epoch 2/10\n", - "512/512 [==============================] - 0s 53us/sample - loss: 1.5484 - accuracy: 0.5625\n", + "523/523 [==============================] - 0s 65us/sample - loss: 1.4263 - accuracy: 0.5813\n", "Epoch 3/10\n", - "512/512 [==============================] - 0s 54us/sample - loss: 1.0831 - accuracy: 0.6309\n", + "523/523 [==============================] - 0s 61us/sample - loss: 1.0224 - accuracy: 0.6750\n", "Epoch 4/10\n", - "512/512 [==============================] - 0s 53us/sample - loss: 0.8084 - accuracy: 0.7383\n", + "523/523 [==============================] - 0s 60us/sample - loss: 0.8609 - accuracy: 0.7228\n", "Epoch 5/10\n", - "512/512 [==============================] - 0s 53us/sample - loss: 0.6601 - accuracy: 0.7734\n", + "523/523 [==============================] - 0s 58us/sample - loss: 0.7089 - accuracy: 0.7514\n", "Epoch 6/10\n", - "512/512 [==============================] - 0s 53us/sample - loss: 0.5029 - accuracy: 0.8398\n", + "523/523 [==============================] - 0s 60us/sample - loss: 0.5944 - accuracy: 0.8031\n", "Epoch 7/10\n", - "512/512 [==============================] - 0s 55us/sample - loss: 0.4373 - accuracy: 0.8555\n", + "523/523 [==============================] - 0s 61us/sample - loss: 0.5785 - accuracy: 0.8164\n", "Epoch 8/10\n", - "512/512 [==============================] - 0s 54us/sample - loss: 0.4090 - accuracy: 0.8652\n", + "523/523 [==============================] - 0s 60us/sample - loss: 0.4617 - accuracy: 0.8489\n", "Epoch 9/10\n", - "512/512 [==============================] - 0s 55us/sample - loss: 0.3498 - accuracy: 0.8809\n", + "523/523 [==============================] - 0s 58us/sample - loss: 0.4921 - accuracy: 0.8394\n", "Epoch 10/10\n", - "512/512 [==============================] - 0s 53us/sample - loss: 0.2817 - accuracy: 0.9238\n", - "Train on 484 samples\n", + "523/523 [==============================] - 0s 60us/sample - loss: 0.3318 - accuracy: 0.8987\n", + "Train on 529 samples\n", "Epoch 1/10\n", - "484/484 [==============================] - 10s 20ms/sample - loss: 2.1475 - accuracy: 0.2500\n", + "529/529 [==============================] - 10s 18ms/sample - loss: 2.1437 - accuracy: 0.2571\n", "Epoch 2/10\n", - "484/484 [==============================] - 0s 60us/sample - loss: 1.4527 - accuracy: 0.5868\n", + "529/529 [==============================] - 0s 64us/sample - loss: 1.4258 - accuracy: 0.5898\n", "Epoch 3/10\n", - "484/484 [==============================] - 0s 55us/sample - loss: 0.9352 - accuracy: 0.7397\n", + "529/529 [==============================] - 0s 62us/sample - loss: 0.9738 - accuracy: 0.7089\n", "Epoch 4/10\n", - "484/484 [==============================] - 0s 55us/sample - loss: 0.7173 - accuracy: 0.7479\n", + "529/529 [==============================] - 0s 62us/sample - loss: 0.7705 - accuracy: 0.7467\n", "Epoch 5/10\n", - "484/484 [==============================] - 0s 55us/sample - loss: 0.5807 - accuracy: 0.8182\n", + "529/529 [==============================] - 0s 60us/sample - loss: 0.6301 - accuracy: 0.8204\n", "Epoch 6/10\n", - "484/484 [==============================] - 0s 54us/sample - loss: 0.4414 - accuracy: 0.8533\n", + "529/529 [==============================] - 0s 61us/sample - loss: 0.4969 - accuracy: 0.8526\n", "Epoch 7/10\n", - "484/484 [==============================] - 0s 55us/sample - loss: 0.4306 - accuracy: 0.8554\n", + "529/529 [==============================] - 0s 62us/sample - loss: 0.4833 - accuracy: 0.8318\n", "Epoch 8/10\n", - "484/484 [==============================] - 0s 55us/sample - loss: 0.2938 - accuracy: 0.9008\n", + "529/529 [==============================] - 0s 60us/sample - loss: 0.3817 - accuracy: 0.8866\n", "Epoch 9/10\n", - "484/484 [==============================] - 0s 53us/sample - loss: 0.2370 - accuracy: 0.9194\n", + "529/529 [==============================] - 0s 60us/sample - loss: 0.3641 - accuracy: 0.8904\n", "Epoch 10/10\n", - "484/484 [==============================] - 0s 55us/sample - loss: 0.2418 - accuracy: 0.9132\n", - "Train on 520 samples\n", + "529/529 [==============================] - 0s 59us/sample - loss: 0.2904 - accuracy: 0.9055\n", + "Train on 528 samples\n", "Epoch 1/10\n", - "520/520 [==============================] - 10s 18ms/sample - loss: 2.1393 - accuracy: 0.2673\n", + "528/528 [==============================] - 1s 1ms/sample - loss: 2.1489 - accuracy: 0.2519\n", "Epoch 2/10\n", - "520/520 [==============================] - 0s 66us/sample - loss: 1.4975 - accuracy: 0.5635\n", + "528/528 [==============================] - 0s 59us/sample - loss: 1.5087 - accuracy: 0.5265\n", "Epoch 3/10\n", - "520/520 [==============================] - 0s 58us/sample - loss: 1.0920 - accuracy: 0.6385\n", + "528/528 [==============================] - 0s 61us/sample - loss: 1.0105 - accuracy: 0.6610\n", "Epoch 4/10\n" ] }, @@ -924,167 +952,167 @@ "name": "stdout", "output_type": "stream", "text": [ - "520/520 [==============================] - 0s 59us/sample - loss: 0.8937 - accuracy: 0.7058\n", + "528/528 [==============================] - 0s 60us/sample - loss: 0.8397 - accuracy: 0.7273\n", "Epoch 5/10\n", - "520/520 [==============================] - 0s 66us/sample - loss: 0.7529 - accuracy: 0.7269\n", + "528/528 [==============================] - 0s 62us/sample - loss: 0.7018 - accuracy: 0.7386\n", "Epoch 6/10\n", - "520/520 [==============================] - 0s 59us/sample - loss: 0.6097 - accuracy: 0.7962\n", + "528/528 [==============================] - 0s 60us/sample - loss: 0.6145 - accuracy: 0.7973\n", "Epoch 7/10\n", - "520/520 [==============================] - 0s 62us/sample - loss: 0.5666 - accuracy: 0.8096\n", + "528/528 [==============================] - 0s 58us/sample - loss: 0.5095 - accuracy: 0.8371\n", "Epoch 8/10\n", - "520/520 [==============================] - 0s 60us/sample - loss: 0.4816 - accuracy: 0.8481\n", + "528/528 [==============================] - 0s 59us/sample - loss: 0.4645 - accuracy: 0.8295\n", "Epoch 9/10\n", - "520/520 [==============================] - 0s 58us/sample - loss: 0.3869 - accuracy: 0.8673\n", + "528/528 [==============================] - 0s 60us/sample - loss: 0.4049 - accuracy: 0.8712\n", "Epoch 10/10\n", - "520/520 [==============================] - 0s 60us/sample - loss: 0.3678 - accuracy: 0.8808\n", - "Train on 503 samples\n", + "528/528 [==============================] - 0s 57us/sample - loss: 0.3533 - accuracy: 0.8864\n", + "Train on 505 samples\n", "Epoch 1/10\n", - "503/503 [==============================] - 10s 19ms/sample - loss: 2.1470 - accuracy: 0.2903\n", + "505/505 [==============================] - 10s 20ms/sample - loss: 2.1431 - accuracy: 0.2337\n", "Epoch 2/10\n", - "503/503 [==============================] - 0s 57us/sample - loss: 1.4585 - accuracy: 0.5964\n", + "505/505 [==============================] - 0s 58us/sample - loss: 1.5446 - accuracy: 0.5564\n", "Epoch 3/10\n", - "503/503 [==============================] - 0s 55us/sample - loss: 0.9445 - accuracy: 0.6899\n", + "505/505 [==============================] - 0s 55us/sample - loss: 1.0494 - accuracy: 0.6950\n", "Epoch 4/10\n", - "503/503 [==============================] - 0s 55us/sample - loss: 0.6983 - accuracy: 0.7435\n", + "505/505 [==============================] - 0s 58us/sample - loss: 0.8355 - accuracy: 0.7287\n", "Epoch 5/10\n", - "503/503 [==============================] - 0s 54us/sample - loss: 0.6330 - accuracy: 0.7714\n", + "505/505 [==============================] - 0s 60us/sample - loss: 0.6153 - accuracy: 0.8119\n", "Epoch 6/10\n", - "503/503 [==============================] - 0s 53us/sample - loss: 0.4556 - accuracy: 0.8449\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.5346 - accuracy: 0.8297\n", "Epoch 7/10\n", - "503/503 [==============================] - 0s 54us/sample - loss: 0.3783 - accuracy: 0.8787\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.4651 - accuracy: 0.8554\n", "Epoch 8/10\n", - "503/503 [==============================] - 0s 52us/sample - loss: 0.2890 - accuracy: 0.9085\n", + "505/505 [==============================] - 0s 56us/sample - loss: 0.3967 - accuracy: 0.8515\n", "Epoch 9/10\n", - "503/503 [==============================] - 0s 54us/sample - loss: 0.2801 - accuracy: 0.9066\n", + "505/505 [==============================] - 0s 56us/sample - loss: 0.3255 - accuracy: 0.8911\n", "Epoch 10/10\n", - "503/503 [==============================] - 0s 54us/sample - loss: 0.2247 - accuracy: 0.9225\n", - "Train on 476 samples\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.2694 - accuracy: 0.9149\n", + "Train on 480 samples\n", "Epoch 1/10\n", - "476/476 [==============================] - 1s 1ms/sample - loss: 2.1596 - accuracy: 0.2395\n", + "480/480 [==============================] - 10s 21ms/sample - loss: 2.1452 - accuracy: 0.2271\n", "Epoch 2/10\n", - "476/476 [==============================] - 0s 55us/sample - loss: 1.5531 - accuracy: 0.5777\n", + "480/480 [==============================] - 0s 59us/sample - loss: 1.4709 - accuracy: 0.6292\n", "Epoch 3/10\n", - "476/476 [==============================] - 0s 55us/sample - loss: 1.1193 - accuracy: 0.6450\n", + "480/480 [==============================] - 0s 57us/sample - loss: 0.8874 - accuracy: 0.7333\n", "Epoch 4/10\n", - "476/476 [==============================] - 0s 55us/sample - loss: 0.8334 - accuracy: 0.7164\n", + "480/480 [==============================] - 0s 57us/sample - loss: 0.6707 - accuracy: 0.7771\n", "Epoch 5/10\n", - "476/476 [==============================] - 0s 62us/sample - loss: 0.6481 - accuracy: 0.7857\n", + "480/480 [==============================] - 0s 55us/sample - loss: 0.5801 - accuracy: 0.8083\n", "Epoch 6/10\n", - "476/476 [==============================] - 0s 54us/sample - loss: 0.5894 - accuracy: 0.8088\n", + "480/480 [==============================] - 0s 56us/sample - loss: 0.4870 - accuracy: 0.8542\n", "Epoch 7/10\n", - "476/476 [==============================] - 0s 54us/sample - loss: 0.4470 - accuracy: 0.8592\n", + "480/480 [==============================] - 0s 63us/sample - loss: 0.3774 - accuracy: 0.8667\n", "Epoch 8/10\n", - "476/476 [==============================] - 0s 55us/sample - loss: 0.4215 - accuracy: 0.8634\n", + "480/480 [==============================] - 0s 57us/sample - loss: 0.3044 - accuracy: 0.9083\n", "Epoch 9/10\n", - "476/476 [==============================] - 0s 56us/sample - loss: 0.3499 - accuracy: 0.8908\n", + "480/480 [==============================] - 0s 55us/sample - loss: 0.2990 - accuracy: 0.8958\n", "Epoch 10/10\n", - "476/476 [==============================] - 0s 54us/sample - loss: 0.2888 - accuracy: 0.9181\n", - "Train on 508 samples\n", + "480/480 [==============================] - 0s 56us/sample - loss: 0.2409 - accuracy: 0.9146\n", + "Train on 468 samples\n", "Epoch 1/10\n", - "508/508 [==============================] - 10s 19ms/sample - loss: 2.1688 - accuracy: 0.2382\n", + "468/468 [==============================] - 10s 21ms/sample - loss: 2.1546 - accuracy: 0.2799\n", "Epoch 2/10\n", - "508/508 [==============================] - 0s 59us/sample - loss: 1.5726 - accuracy: 0.5551\n", + "468/468 [==============================] - 0s 60us/sample - loss: 1.5467 - accuracy: 0.5449\n", "Epoch 3/10\n", - "508/508 [==============================] - 0s 56us/sample - loss: 1.0846 - accuracy: 0.6634\n", + "468/468 [==============================] - 0s 57us/sample - loss: 0.9986 - accuracy: 0.6838\n", "Epoch 4/10\n", - "508/508 [==============================] - 0s 54us/sample - loss: 0.8461 - accuracy: 0.7402\n", + "468/468 [==============================] - 0s 57us/sample - loss: 0.7478 - accuracy: 0.7585\n", "Epoch 5/10\n", - "508/508 [==============================] - 0s 53us/sample - loss: 0.6267 - accuracy: 0.8012\n", + "468/468 [==============================] - 0s 57us/sample - loss: 0.6031 - accuracy: 0.7991\n", "Epoch 6/10\n", - "508/508 [==============================] - 0s 54us/sample - loss: 0.5296 - accuracy: 0.8228\n", + "468/468 [==============================] - 0s 57us/sample - loss: 0.4572 - accuracy: 0.8654\n", "Epoch 7/10\n", - "508/508 [==============================] - 0s 53us/sample - loss: 0.4193 - accuracy: 0.8681\n", + "468/468 [==============================] - 0s 91us/sample - loss: 0.4436 - accuracy: 0.8654\n", "Epoch 8/10\n", - "508/508 [==============================] - 0s 54us/sample - loss: 0.3688 - accuracy: 0.8937\n", + "468/468 [==============================] - 0s 57us/sample - loss: 0.3346 - accuracy: 0.9017\n", "Epoch 9/10\n", - "508/508 [==============================] - 0s 53us/sample - loss: 0.3436 - accuracy: 0.8858\n", + "468/468 [==============================] - 0s 59us/sample - loss: 0.3536 - accuracy: 0.8761\n", "Epoch 10/10\n", - "508/508 [==============================] - 0s 54us/sample - loss: 0.2736 - accuracy: 0.9154\n", + "468/468 [==============================] - 0s 58us/sample - loss: 0.2524 - accuracy: 0.9103\n", "Train on 486 samples\n", "Epoch 1/10\n", - "486/486 [==============================] - 10s 20ms/sample - loss: 2.1343 - accuracy: 0.2551\n", + "486/486 [==============================] - 10s 21ms/sample - loss: 2.1433 - accuracy: 0.2551\n", "Epoch 2/10\n", - "486/486 [==============================] - 0s 59us/sample - loss: 1.5164 - accuracy: 0.5741\n", + "486/486 [==============================] - 0s 60us/sample - loss: 1.4641 - accuracy: 0.5720\n", "Epoch 3/10\n", - "486/486 [==============================] - 0s 56us/sample - loss: 1.0556 - accuracy: 0.6687\n", + "486/486 [==============================] - 0s 77us/sample - loss: 1.0028 - accuracy: 0.6770\n", "Epoch 4/10\n", - "486/486 [==============================] - 0s 56us/sample - loss: 0.7575 - accuracy: 0.7737\n", + "486/486 [==============================] - 0s 56us/sample - loss: 0.8034 - accuracy: 0.7366\n", "Epoch 5/10\n", - "486/486 [==============================] - 0s 57us/sample - loss: 0.5977 - accuracy: 0.8107\n", + "486/486 [==============================] - 0s 56us/sample - loss: 0.6493 - accuracy: 0.7901\n", "Epoch 6/10\n", - "486/486 [==============================] - 0s 60us/sample - loss: 0.5502 - accuracy: 0.8210\n", + "486/486 [==============================] - 0s 57us/sample - loss: 0.5624 - accuracy: 0.8086\n", "Epoch 7/10\n", - "486/486 [==============================] - 0s 57us/sample - loss: 0.4591 - accuracy: 0.8519\n", + "486/486 [==============================] - 0s 55us/sample - loss: 0.5094 - accuracy: 0.8395\n", "Epoch 8/10\n", - "486/486 [==============================] - 0s 56us/sample - loss: 0.3858 - accuracy: 0.8951\n", + "486/486 [==============================] - 0s 57us/sample - loss: 0.4072 - accuracy: 0.8704\n", "Epoch 9/10\n", - "486/486 [==============================] - 0s 56us/sample - loss: 0.3276 - accuracy: 0.8827\n", + "486/486 [==============================] - 0s 55us/sample - loss: 0.3726 - accuracy: 0.8807\n", "Epoch 10/10\n", - "486/486 [==============================] - 0s 57us/sample - loss: 0.2722 - accuracy: 0.9218\n", - "Train on 488 samples\n", + "486/486 [==============================] - 0s 56us/sample - loss: 0.2992 - accuracy: 0.9095\n", + "Train on 495 samples\n", "Epoch 1/10\n", - "488/488 [==============================] - 10s 20ms/sample - loss: 2.1683 - accuracy: 0.2336\n", + "495/495 [==============================] - 10s 20ms/sample - loss: 2.1340 - accuracy: 0.2586\n", "Epoch 2/10\n", - "488/488 [==============================] - 0s 58us/sample - loss: 1.4995 - accuracy: 0.5840\n", + "495/495 [==============================] - 0s 60us/sample - loss: 1.5234 - accuracy: 0.5737\n", "Epoch 3/10\n", - "488/488 [==============================] - 0s 56us/sample - loss: 1.0556 - accuracy: 0.6332\n", + "495/495 [==============================] - 0s 56us/sample - loss: 1.0101 - accuracy: 0.6727\n", "Epoch 4/10\n", - "488/488 [==============================] - 0s 54us/sample - loss: 0.8058 - accuracy: 0.7213\n", + "495/495 [==============================] - 0s 56us/sample - loss: 0.8037 - accuracy: 0.7313\n", "Epoch 5/10\n", - "488/488 [==============================] - 0s 54us/sample - loss: 0.6348 - accuracy: 0.7828\n", + "495/495 [==============================] - 0s 56us/sample - loss: 0.6052 - accuracy: 0.8040\n", "Epoch 6/10\n", - "488/488 [==============================] - 0s 54us/sample - loss: 0.5106 - accuracy: 0.8217\n", + "495/495 [==============================] - 0s 57us/sample - loss: 0.5539 - accuracy: 0.8424\n", "Epoch 7/10\n", - "488/488 [==============================] - 0s 54us/sample - loss: 0.4071 - accuracy: 0.8689\n", + "495/495 [==============================] - 0s 61us/sample - loss: 0.4627 - accuracy: 0.8586\n", "Epoch 8/10\n", - "488/488 [==============================] - 0s 55us/sample - loss: 0.3513 - accuracy: 0.8996\n", + "495/495 [==============================] - 0s 60us/sample - loss: 0.4110 - accuracy: 0.8727\n", "Epoch 9/10\n", - "488/488 [==============================] - 0s 53us/sample - loss: 0.2743 - accuracy: 0.9221\n", + "495/495 [==============================] - 0s 61us/sample - loss: 0.3609 - accuracy: 0.8869\n", "Epoch 10/10\n", - "488/488 [==============================] - 0s 61us/sample - loss: 0.2793 - accuracy: 0.8975\n", - "Train on 480 samples\n", + "495/495 [==============================] - 0s 56us/sample - loss: 0.3141 - accuracy: 0.9091\n", + "Train on 482 samples\n", "Epoch 1/10\n", - "480/480 [==============================] - 10s 20ms/sample - loss: 2.1472 - accuracy: 0.2604\n", + "482/482 [==============================] - 10s 21ms/sample - loss: 2.1764 - accuracy: 0.2303\n", "Epoch 2/10\n", - "480/480 [==============================] - 0s 60us/sample - loss: 1.4224 - accuracy: 0.6438\n", + "482/482 [==============================] - 0s 60us/sample - loss: 1.5406 - accuracy: 0.5560\n", "Epoch 3/10\n", - "480/480 [==============================] - 0s 53us/sample - loss: 0.8888 - accuracy: 0.7000\n", + "482/482 [==============================] - 0s 57us/sample - loss: 1.0473 - accuracy: 0.6784\n", "Epoch 4/10\n", - "480/480 [==============================] - 0s 55us/sample - loss: 0.6991 - accuracy: 0.7812\n", + "482/482 [==============================] - 0s 56us/sample - loss: 0.8430 - accuracy: 0.7324\n", "Epoch 5/10\n", - "480/480 [==============================] - 0s 55us/sample - loss: 0.5537 - accuracy: 0.8104\n", + "482/482 [==============================] - 0s 56us/sample - loss: 0.6383 - accuracy: 0.8112\n", "Epoch 6/10\n", - "480/480 [==============================] - 0s 55us/sample - loss: 0.4511 - accuracy: 0.8708\n", + "482/482 [==============================] - 0s 64us/sample - loss: 0.5764 - accuracy: 0.8257\n", "Epoch 7/10\n", - "480/480 [==============================] - 0s 55us/sample - loss: 0.3527 - accuracy: 0.9021\n", + "482/482 [==============================] - 0s 58us/sample - loss: 0.4598 - accuracy: 0.8755\n", "Epoch 8/10\n", - "480/480 [==============================] - 0s 55us/sample - loss: 0.2919 - accuracy: 0.9167\n", + "482/482 [==============================] - 0s 59us/sample - loss: 0.3815 - accuracy: 0.8921\n", "Epoch 9/10\n", - "480/480 [==============================] - 0s 54us/sample - loss: 0.2120 - accuracy: 0.9354\n", + "482/482 [==============================] - 0s 57us/sample - loss: 0.3297 - accuracy: 0.9046\n", "Epoch 10/10\n", - "480/480 [==============================] - 0s 54us/sample - loss: 0.2068 - accuracy: 0.9417\n", - "Train on 498 samples\n", + "482/482 [==============================] - 0s 59us/sample - loss: 0.2295 - accuracy: 0.9357\n", + "Train on 486 samples\n", "Epoch 1/10\n", - "498/498 [==============================] - 10s 20ms/sample - loss: 2.1714 - accuracy: 0.2329\n", + "486/486 [==============================] - 1s 1ms/sample - loss: 2.1878 - accuracy: 0.2366\n", "Epoch 2/10\n", - "498/498 [==============================] - 0s 75us/sample - loss: 1.5817 - accuracy: 0.5221\n", + "486/486 [==============================] - 0s 56us/sample - loss: 1.5752 - accuracy: 0.6214\n", "Epoch 3/10\n", - "498/498 [==============================] - 0s 56us/sample - loss: 1.1751 - accuracy: 0.6145\n", + "486/486 [==============================] - 0s 56us/sample - loss: 1.0463 - accuracy: 0.6708\n", "Epoch 4/10\n", - "498/498 [==============================] - 0s 57us/sample - loss: 0.7997 - accuracy: 0.7610\n", + "486/486 [==============================] - 0s 55us/sample - loss: 0.8246 - accuracy: 0.7366\n", "Epoch 5/10\n", - "498/498 [==============================] - 0s 56us/sample - loss: 0.7069 - accuracy: 0.7791\n", + "486/486 [==============================] - 0s 85us/sample - loss: 0.6613 - accuracy: 0.7757\n", "Epoch 6/10\n", - "498/498 [==============================] - 0s 55us/sample - loss: 0.5499 - accuracy: 0.8273\n", + "486/486 [==============================] - 0s 57us/sample - loss: 0.5554 - accuracy: 0.8189\n", "Epoch 7/10\n", - "498/498 [==============================] - 0s 54us/sample - loss: 0.4973 - accuracy: 0.8474\n", + "486/486 [==============================] - 0s 56us/sample - loss: 0.4802 - accuracy: 0.8395\n", "Epoch 8/10\n", - "498/498 [==============================] - 0s 58us/sample - loss: 0.4175 - accuracy: 0.8614\n", + "486/486 [==============================] - 0s 55us/sample - loss: 0.3758 - accuracy: 0.8827\n", "Epoch 9/10\n", - "498/498 [==============================] - 0s 54us/sample - loss: 0.3616 - accuracy: 0.8775\n", + "486/486 [==============================] - 0s 55us/sample - loss: 0.3418 - accuracy: 0.8930\n", "Epoch 10/10\n", - "498/498 [==============================] - 0s 54us/sample - loss: 0.3297 - accuracy: 0.8896\n", - "Train on 476 samples\n", + "486/486 [==============================] - 0s 56us/sample - loss: 0.2939 - accuracy: 0.9012\n", + "Train on 534 samples\n", "Epoch 1/10\n" ] }, @@ -1092,168 +1120,168 @@ "name": "stdout", "output_type": "stream", "text": [ - "476/476 [==============================] - 1s 1ms/sample - loss: 2.1658 - accuracy: 0.2605\n", + "534/534 [==============================] - 10s 19ms/sample - loss: 2.1621 - accuracy: 0.2154\n", "Epoch 2/10\n", - "476/476 [==============================] - 0s 88us/sample - loss: 1.5388 - accuracy: 0.5399\n", + "534/534 [==============================] - 0s 70us/sample - loss: 1.5126 - accuracy: 0.5056\n", "Epoch 3/10\n", - "476/476 [==============================] - 0s 60us/sample - loss: 1.0840 - accuracy: 0.6450\n", + "534/534 [==============================] - 0s 64us/sample - loss: 1.0817 - accuracy: 0.6217\n", "Epoch 4/10\n", - "476/476 [==============================] - 0s 55us/sample - loss: 0.8079 - accuracy: 0.7269\n", + "534/534 [==============================] - 0s 63us/sample - loss: 0.7903 - accuracy: 0.7509\n", "Epoch 5/10\n", - "476/476 [==============================] - 0s 56us/sample - loss: 0.6481 - accuracy: 0.7773\n", + "534/534 [==============================] - 0s 60us/sample - loss: 0.6815 - accuracy: 0.7884\n", "Epoch 6/10\n", - "476/476 [==============================] - 0s 54us/sample - loss: 0.4718 - accuracy: 0.8466\n", + "534/534 [==============================] - 0s 60us/sample - loss: 0.5505 - accuracy: 0.8296\n", "Epoch 7/10\n", - "476/476 [==============================] - 0s 56us/sample - loss: 0.4352 - accuracy: 0.8550\n", + "534/534 [==============================] - 0s 59us/sample - loss: 0.5044 - accuracy: 0.8446\n", "Epoch 8/10\n", - "476/476 [==============================] - 0s 62us/sample - loss: 0.3376 - accuracy: 0.8845\n", + "534/534 [==============================] - 0s 59us/sample - loss: 0.4086 - accuracy: 0.8727\n", "Epoch 9/10\n", - "476/476 [==============================] - 0s 57us/sample - loss: 0.3099 - accuracy: 0.9013\n", + "534/534 [==============================] - 0s 59us/sample - loss: 0.3429 - accuracy: 0.9064\n", "Epoch 10/10\n", - "476/476 [==============================] - 0s 56us/sample - loss: 0.3038 - accuracy: 0.9076\n", - "Train on 499 samples\n", + "534/534 [==============================] - 0s 58us/sample - loss: 0.3056 - accuracy: 0.8970\n", + "Train on 503 samples\n", "Epoch 1/10\n", - "499/499 [==============================] - 10s 20ms/sample - loss: 2.1521 - accuracy: 0.2525\n", + "503/503 [==============================] - 10s 21ms/sample - loss: 2.1127 - accuracy: 0.3419\n", "Epoch 2/10\n", - "499/499 [==============================] - 0s 57us/sample - loss: 1.5282 - accuracy: 0.5772\n", + "503/503 [==============================] - 0s 60us/sample - loss: 1.4300 - accuracy: 0.5905\n", "Epoch 3/10\n", - "499/499 [==============================] - 0s 57us/sample - loss: 0.9815 - accuracy: 0.7114\n", + "503/503 [==============================] - 0s 56us/sample - loss: 0.9352 - accuracy: 0.7177\n", "Epoch 4/10\n", - "499/499 [==============================] - 0s 55us/sample - loss: 0.7844 - accuracy: 0.7575\n", + "503/503 [==============================] - 0s 56us/sample - loss: 0.7279 - accuracy: 0.7594\n", "Epoch 5/10\n", - "499/499 [==============================] - 0s 55us/sample - loss: 0.5814 - accuracy: 0.8337\n", + "503/503 [==============================] - 0s 57us/sample - loss: 0.5622 - accuracy: 0.8270\n", "Epoch 6/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 0.5117 - accuracy: 0.8357\n", + "503/503 [==============================] - 0s 56us/sample - loss: 0.4606 - accuracy: 0.8390\n", "Epoch 7/10\n", - "499/499 [==============================] - 0s 55us/sample - loss: 0.4182 - accuracy: 0.8737\n", + "503/503 [==============================] - 0s 55us/sample - loss: 0.3950 - accuracy: 0.8569\n", "Epoch 8/10\n", - "499/499 [==============================] - 0s 56us/sample - loss: 0.3845 - accuracy: 0.8597\n", + "503/503 [==============================] - 0s 55us/sample - loss: 0.3185 - accuracy: 0.9026\n", "Epoch 9/10\n", - "499/499 [==============================] - 0s 55us/sample - loss: 0.3050 - accuracy: 0.9058\n", + "503/503 [==============================] - 0s 62us/sample - loss: 0.2429 - accuracy: 0.9264\n", "Epoch 10/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 0.2917 - accuracy: 0.9038\n", - "Train on 489 samples\n", + "503/503 [==============================] - 0s 55us/sample - loss: 0.2266 - accuracy: 0.9245\n", + "Train on 492 samples\n", "Epoch 1/10\n", - "489/489 [==============================] - 10s 21ms/sample - loss: 2.1476 - accuracy: 0.2761\n", + "492/492 [==============================] - 10s 21ms/sample - loss: 2.1539 - accuracy: 0.2927\n", "Epoch 2/10\n", - "489/489 [==============================] - 0s 59us/sample - loss: 1.4879 - accuracy: 0.6033\n", + "492/492 [==============================] - 0s 59us/sample - loss: 1.4890 - accuracy: 0.6504\n", "Epoch 3/10\n", - "489/489 [==============================] - 0s 55us/sample - loss: 0.9655 - accuracy: 0.7198\n", + "492/492 [==============================] - 0s 56us/sample - loss: 0.9864 - accuracy: 0.7093\n", "Epoch 4/10\n", - "489/489 [==============================] - 0s 64us/sample - loss: 0.7375 - accuracy: 0.7505\n", + "492/492 [==============================] - 0s 57us/sample - loss: 0.6970 - accuracy: 0.7683\n", "Epoch 5/10\n", - "489/489 [==============================] - 0s 55us/sample - loss: 0.5713 - accuracy: 0.8180\n", + "492/492 [==============================] - 0s 56us/sample - loss: 0.5769 - accuracy: 0.8110\n", "Epoch 6/10\n", - "489/489 [==============================] - 0s 56us/sample - loss: 0.4545 - accuracy: 0.8671\n", + "492/492 [==============================] - 0s 56us/sample - loss: 0.4525 - accuracy: 0.8720\n", "Epoch 7/10\n", - "489/489 [==============================] - 0s 55us/sample - loss: 0.4251 - accuracy: 0.8732\n", + "492/492 [==============================] - 0s 55us/sample - loss: 0.4082 - accuracy: 0.8598\n", "Epoch 8/10\n", - "489/489 [==============================] - 0s 55us/sample - loss: 0.3244 - accuracy: 0.9059\n", + "492/492 [==============================] - 0s 56us/sample - loss: 0.3098 - accuracy: 0.8984\n", "Epoch 9/10\n", - "489/489 [==============================] - 0s 55us/sample - loss: 0.3131 - accuracy: 0.9080\n", + "492/492 [==============================] - 0s 56us/sample - loss: 0.3037 - accuracy: 0.8943\n", "Epoch 10/10\n", - "489/489 [==============================] - 0s 55us/sample - loss: 0.2341 - accuracy: 0.9264\n", - "Train on 524 samples\n", + "492/492 [==============================] - 0s 56us/sample - loss: 0.2356 - accuracy: 0.9167\n", + "Train on 500 samples\n", "Epoch 1/10\n", - "524/524 [==============================] - 10s 19ms/sample - loss: 2.1665 - accuracy: 0.2195\n", + "500/500 [==============================] - 11s 21ms/sample - loss: 2.1650 - accuracy: 0.2420\n", "Epoch 2/10\n", - "524/524 [==============================] - 0s 62us/sample - loss: 1.5493 - accuracy: 0.5210\n", + "500/500 [==============================] - 0s 59us/sample - loss: 1.5340 - accuracy: 0.5620\n", "Epoch 3/10\n", - "524/524 [==============================] - 0s 64us/sample - loss: 1.0477 - accuracy: 0.6622\n", + "500/500 [==============================] - 0s 56us/sample - loss: 1.0365 - accuracy: 0.6600\n", "Epoch 4/10\n", - "524/524 [==============================] - 0s 61us/sample - loss: 0.8519 - accuracy: 0.7271\n", + "500/500 [==============================] - 0s 56us/sample - loss: 0.8003 - accuracy: 0.7240\n", "Epoch 5/10\n", - "524/524 [==============================] - 0s 60us/sample - loss: 0.7305 - accuracy: 0.7557\n", + "500/500 [==============================] - 0s 57us/sample - loss: 0.6125 - accuracy: 0.7860\n", "Epoch 6/10\n", - "524/524 [==============================] - 0s 62us/sample - loss: 0.6277 - accuracy: 0.7748\n", + "500/500 [==============================] - 0s 57us/sample - loss: 0.5069 - accuracy: 0.8460\n", "Epoch 7/10\n", - "524/524 [==============================] - 0s 62us/sample - loss: 0.5449 - accuracy: 0.7996\n", + "500/500 [==============================] - 0s 57us/sample - loss: 0.4328 - accuracy: 0.8720\n", "Epoch 8/10\n", - "524/524 [==============================] - 0s 60us/sample - loss: 0.4876 - accuracy: 0.8263\n", + "500/500 [==============================] - 0s 57us/sample - loss: 0.3756 - accuracy: 0.8720\n", "Epoch 9/10\n", - "524/524 [==============================] - 0s 60us/sample - loss: 0.4214 - accuracy: 0.8664\n", + "500/500 [==============================] - 0s 57us/sample - loss: 0.3358 - accuracy: 0.8780\n", "Epoch 10/10\n", - "524/524 [==============================] - 0s 57us/sample - loss: 0.3627 - accuracy: 0.8874\n", - "Train on 499 samples\n", + "500/500 [==============================] - 0s 56us/sample - loss: 0.2677 - accuracy: 0.9120\n", + "Train on 488 samples\n", "Epoch 1/10\n", - "499/499 [==============================] - 1s 1ms/sample - loss: 2.1702 - accuracy: 0.2405\n", + "488/488 [==============================] - 1s 1ms/sample - loss: 2.1246 - accuracy: 0.2807\n", "Epoch 2/10\n", - "499/499 [==============================] - 0s 55us/sample - loss: 1.5764 - accuracy: 0.5912\n", + "488/488 [==============================] - 0s 56us/sample - loss: 1.4881 - accuracy: 0.5615\n", "Epoch 3/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 1.0004 - accuracy: 0.7014\n", + "488/488 [==============================] - 0s 56us/sample - loss: 0.9723 - accuracy: 0.6988\n", "Epoch 4/10\n", - "499/499 [==============================] - 0s 55us/sample - loss: 0.6617 - accuracy: 0.7856\n", + "488/488 [==============================] - 0s 73us/sample - loss: 0.7687 - accuracy: 0.7275\n", "Epoch 5/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 0.5824 - accuracy: 0.8257\n", + "488/488 [==============================] - 0s 56us/sample - loss: 0.5672 - accuracy: 0.8258\n", "Epoch 6/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 0.4763 - accuracy: 0.8377\n", + "488/488 [==============================] - 0s 58us/sample - loss: 0.4588 - accuracy: 0.8566\n", "Epoch 7/10\n", - "499/499 [==============================] - 0s 53us/sample - loss: 0.4000 - accuracy: 0.8597\n", + "488/488 [==============================] - 0s 58us/sample - loss: 0.4122 - accuracy: 0.8504\n", "Epoch 8/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 0.3440 - accuracy: 0.8818\n", + "488/488 [==============================] - 0s 61us/sample - loss: 0.3228 - accuracy: 0.8975\n", "Epoch 9/10\n", - "499/499 [==============================] - 0s 54us/sample - loss: 0.3045 - accuracy: 0.9178\n", + "488/488 [==============================] - 0s 57us/sample - loss: 0.3162 - accuracy: 0.9098\n", "Epoch 10/10\n", - "499/499 [==============================] - 0s 56us/sample - loss: 0.2427 - accuracy: 0.9078\n", - "Train on 520 samples\n", + "488/488 [==============================] - 0s 57us/sample - loss: 0.2826 - accuracy: 0.9139\n", + "Train on 518 samples\n", "Epoch 1/10\n", - "520/520 [==============================] - 1s 1ms/sample - loss: 2.1108 - accuracy: 0.2904\n", + "518/518 [==============================] - 11s 20ms/sample - loss: 2.1700 - accuracy: 0.2220\n", "Epoch 2/10\n", - "520/520 [==============================] - 0s 61us/sample - loss: 1.4075 - accuracy: 0.5769\n", + "518/518 [==============================] - 0s 64us/sample - loss: 1.5355 - accuracy: 0.5579\n", "Epoch 3/10\n", - "520/520 [==============================] - 0s 61us/sample - loss: 1.0159 - accuracy: 0.6538\n", + "518/518 [==============================] - 0s 62us/sample - loss: 1.0461 - accuracy: 0.6757\n", "Epoch 4/10\n", - "520/520 [==============================] - 0s 61us/sample - loss: 0.8287 - accuracy: 0.7462\n", + "518/518 [==============================] - 0s 113us/sample - loss: 0.8217 - accuracy: 0.7181\n", "Epoch 5/10\n", - "520/520 [==============================] - 0s 61us/sample - loss: 0.6397 - accuracy: 0.7769\n", + "518/518 [==============================] - 0s 62us/sample - loss: 0.7038 - accuracy: 0.7625\n", "Epoch 6/10\n", - "520/520 [==============================] - 0s 59us/sample - loss: 0.5619 - accuracy: 0.8115\n", + "518/518 [==============================] - 0s 61us/sample - loss: 0.5637 - accuracy: 0.8205\n", "Epoch 7/10\n", - "520/520 [==============================] - 0s 61us/sample - loss: 0.4385 - accuracy: 0.8462\n", + "518/518 [==============================] - 0s 63us/sample - loss: 0.5096 - accuracy: 0.8359\n", "Epoch 8/10\n", - "520/520 [==============================] - 0s 59us/sample - loss: 0.4131 - accuracy: 0.8442\n", + "518/518 [==============================] - 0s 63us/sample - loss: 0.4425 - accuracy: 0.8475\n", "Epoch 9/10\n", - "520/520 [==============================] - 0s 59us/sample - loss: 0.4337 - accuracy: 0.8712\n", + "518/518 [==============================] - 0s 61us/sample - loss: 0.3985 - accuracy: 0.8571\n", "Epoch 10/10\n", - "520/520 [==============================] - 0s 59us/sample - loss: 0.3679 - accuracy: 0.8712\n", - "Train on 527 samples\n", + "518/518 [==============================] - 0s 62us/sample - loss: 0.3981 - accuracy: 0.8784\n", + "Train on 505 samples\n", "Epoch 1/10\n", - "527/527 [==============================] - 10s 20ms/sample - loss: 2.1267 - accuracy: 0.2979\n", + "505/505 [==============================] - 1s 1ms/sample - loss: 2.1348 - accuracy: 0.3030\n", "Epoch 2/10\n", - "527/527 [==============================] - 0s 63us/sample - loss: 1.3446 - accuracy: 0.6186\n", + "505/505 [==============================] - 0s 56us/sample - loss: 1.4776 - accuracy: 0.5584\n", "Epoch 3/10\n", - "527/527 [==============================] - 0s 62us/sample - loss: 1.0363 - accuracy: 0.6641\n", + "505/505 [==============================] - 0s 56us/sample - loss: 0.9705 - accuracy: 0.6832\n", "Epoch 4/10\n", - "527/527 [==============================] - 0s 61us/sample - loss: 0.7397 - accuracy: 0.7628\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.7680 - accuracy: 0.7465\n", "Epoch 5/10\n", - "527/527 [==============================] - 0s 61us/sample - loss: 0.6275 - accuracy: 0.8216\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.6106 - accuracy: 0.8079\n", "Epoch 6/10\n", - "527/527 [==============================] - 0s 61us/sample - loss: 0.5667 - accuracy: 0.8292\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.5002 - accuracy: 0.8218\n", "Epoch 7/10\n", - "527/527 [==============================] - 0s 61us/sample - loss: 0.4883 - accuracy: 0.8444\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.4429 - accuracy: 0.8535\n", "Epoch 8/10\n", - "527/527 [==============================] - 0s 64us/sample - loss: 0.4091 - accuracy: 0.8748\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.3665 - accuracy: 0.8812\n", "Epoch 9/10\n", - "527/527 [==============================] - 0s 60us/sample - loss: 0.3648 - accuracy: 0.8767\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.3601 - accuracy: 0.8851\n", "Epoch 10/10\n", - "527/527 [==============================] - 0s 59us/sample - loss: 0.2951 - accuracy: 0.9032\n", - "Train on 488 samples\n", + "505/505 [==============================] - 0s 55us/sample - loss: 0.3121 - accuracy: 0.8970\n", + "Train on 483 samples\n", "Epoch 1/10\n", - "488/488 [==============================] - 1s 1ms/sample - loss: 2.1751 - accuracy: 0.2828\n", + "483/483 [==============================] - 11s 22ms/sample - loss: 2.1250 - accuracy: 0.2567\n", "Epoch 2/10\n", - "488/488 [==============================] - 0s 58us/sample - loss: 1.5316 - accuracy: 0.5717\n", + "483/483 [==============================] - 0s 60us/sample - loss: 1.4379 - accuracy: 0.5818\n", "Epoch 3/10\n", - "488/488 [==============================] - 0s 57us/sample - loss: 1.0565 - accuracy: 0.6619\n", + "483/483 [==============================] - 0s 63us/sample - loss: 0.9778 - accuracy: 0.6687\n", "Epoch 4/10\n", - "488/488 [==============================] - 0s 55us/sample - loss: 0.8274 - accuracy: 0.7398\n", + "483/483 [==============================] - 0s 59us/sample - loss: 0.7335 - accuracy: 0.7764\n", "Epoch 5/10\n", - "488/488 [==============================] - 0s 56us/sample - loss: 0.5862 - accuracy: 0.8033\n", + "483/483 [==============================] - 0s 57us/sample - loss: 0.5719 - accuracy: 0.8385\n", "Epoch 6/10\n", - "488/488 [==============================] - 0s 56us/sample - loss: 0.5475 - accuracy: 0.8217\n", + "483/483 [==============================] - 0s 63us/sample - loss: 0.4996 - accuracy: 0.8571\n", "Epoch 7/10\n", - "488/488 [==============================] - 0s 56us/sample - loss: 0.4342 - accuracy: 0.8689\n", + "483/483 [==============================] - 0s 57us/sample - loss: 0.4375 - accuracy: 0.8675\n", "Epoch 8/10\n", - "488/488 [==============================] - 0s 57us/sample - loss: 0.3966 - accuracy: 0.8730\n", + "483/483 [==============================] - 0s 56us/sample - loss: 0.3887 - accuracy: 0.8841\n", "Epoch 9/10\n" ] }, @@ -1261,167 +1289,167 @@ "name": "stdout", "output_type": "stream", "text": [ - "488/488 [==============================] - 0s 55us/sample - loss: 0.3812 - accuracy: 0.8689\n", + "483/483 [==============================] - 0s 55us/sample - loss: 0.3106 - accuracy: 0.9110\n", "Epoch 10/10\n", - "488/488 [==============================] - 0s 55us/sample - loss: 0.2635 - accuracy: 0.9057\n", - "Train on 309 samples\n", + "483/483 [==============================] - 0s 55us/sample - loss: 0.2805 - accuracy: 0.9110\n", + "Train on 324 samples\n", "Epoch 1/10\n", - "309/309 [==============================] - 10s 34ms/sample - loss: 2.2287 - accuracy: 0.1586\n", + "324/324 [==============================] - 11s 33ms/sample - loss: 2.2548 - accuracy: 0.1420\n", "Epoch 2/10\n", - "309/309 [==============================] - 0s 65us/sample - loss: 1.8008 - accuracy: 0.5275\n", + "324/324 [==============================] - 0s 67us/sample - loss: 1.8664 - accuracy: 0.4753\n", "Epoch 3/10\n", - "309/309 [==============================] - 0s 61us/sample - loss: 1.2902 - accuracy: 0.6343\n", + "324/324 [==============================] - 0s 61us/sample - loss: 1.4684 - accuracy: 0.5864\n", "Epoch 4/10\n", - "309/309 [==============================] - 0s 61us/sample - loss: 0.8978 - accuracy: 0.7508\n", + "324/324 [==============================] - 0s 69us/sample - loss: 1.1296 - accuracy: 0.6265\n", "Epoch 5/10\n", - "309/309 [==============================] - 0s 62us/sample - loss: 0.7633 - accuracy: 0.7670\n", + "324/324 [==============================] - 0s 63us/sample - loss: 0.8679 - accuracy: 0.7160\n", "Epoch 6/10\n", - "309/309 [==============================] - 0s 62us/sample - loss: 0.6468 - accuracy: 0.7896\n", + "324/324 [==============================] - 0s 61us/sample - loss: 0.7579 - accuracy: 0.7500\n", "Epoch 7/10\n", - "309/309 [==============================] - 0s 64us/sample - loss: 0.5144 - accuracy: 0.8220\n", + "324/324 [==============================] - 0s 60us/sample - loss: 0.7367 - accuracy: 0.7562\n", "Epoch 8/10\n", - "309/309 [==============================] - 0s 62us/sample - loss: 0.4538 - accuracy: 0.8447\n", + "324/324 [==============================] - 0s 63us/sample - loss: 0.6104 - accuracy: 0.8179\n", "Epoch 9/10\n", - "309/309 [==============================] - 0s 63us/sample - loss: 0.4069 - accuracy: 0.8706\n", + "324/324 [==============================] - 0s 61us/sample - loss: 0.5102 - accuracy: 0.8426\n", "Epoch 10/10\n", - "309/309 [==============================] - 0s 64us/sample - loss: 0.3559 - accuracy: 0.8932\n", - "Train on 344 samples\n", + "324/324 [==============================] - 0s 63us/sample - loss: 0.4051 - accuracy: 0.8488\n", + "Train on 323 samples\n", "Epoch 1/10\n", - "344/344 [==============================] - 11s 31ms/sample - loss: 2.2280 - accuracy: 0.2064\n", + "323/323 [==============================] - 1s 2ms/sample - loss: 2.2306 - accuracy: 0.1734\n", "Epoch 2/10\n", - "344/344 [==============================] - 0s 64us/sample - loss: 1.8158 - accuracy: 0.4767\n", + "323/323 [==============================] - 0s 62us/sample - loss: 1.7384 - accuracy: 0.5480\n", "Epoch 3/10\n", - "344/344 [==============================] - 0s 59us/sample - loss: 1.3537 - accuracy: 0.6017\n", + "323/323 [==============================] - 0s 62us/sample - loss: 1.2520 - accuracy: 0.6563\n", "Epoch 4/10\n", - "344/344 [==============================] - 0s 59us/sample - loss: 0.9780 - accuracy: 0.7093\n", + "323/323 [==============================] - 0s 64us/sample - loss: 0.8496 - accuracy: 0.7492\n", "Epoch 5/10\n", - "344/344 [==============================] - 0s 59us/sample - loss: 0.7357 - accuracy: 0.7674\n", + "323/323 [==============================] - 0s 63us/sample - loss: 0.6620 - accuracy: 0.7709\n", "Epoch 6/10\n", - "344/344 [==============================] - 0s 58us/sample - loss: 0.6547 - accuracy: 0.8023\n", + "323/323 [==============================] - 0s 63us/sample - loss: 0.4981 - accuracy: 0.8483\n", "Epoch 7/10\n", - "344/344 [==============================] - 0s 60us/sample - loss: 0.5221 - accuracy: 0.8372\n", + "323/323 [==============================] - 0s 64us/sample - loss: 0.4409 - accuracy: 0.8390\n", "Epoch 8/10\n", - "344/344 [==============================] - 0s 59us/sample - loss: 0.4772 - accuracy: 0.8692\n", + "323/323 [==============================] - 0s 62us/sample - loss: 0.3703 - accuracy: 0.8607\n", "Epoch 9/10\n", - "344/344 [==============================] - 0s 59us/sample - loss: 0.4037 - accuracy: 0.8721\n", + "323/323 [==============================] - 0s 62us/sample - loss: 0.3235 - accuracy: 0.9040\n", "Epoch 10/10\n", - "344/344 [==============================] - 0s 57us/sample - loss: 0.3460 - accuracy: 0.8895\n", - "Train on 325 samples\n", + "323/323 [==============================] - 0s 64us/sample - loss: 0.2937 - accuracy: 0.9071\n", + "Train on 342 samples\n", "Epoch 1/10\n", - "325/325 [==============================] - 11s 33ms/sample - loss: 2.2302 - accuracy: 0.1600\n", + "342/342 [==============================] - 11s 32ms/sample - loss: 2.2380 - accuracy: 0.1725\n", "Epoch 2/10\n", - "325/325 [==============================] - 0s 68us/sample - loss: 1.8088 - accuracy: 0.4431\n", + "342/342 [==============================] - 0s 65us/sample - loss: 1.8316 - accuracy: 0.4766\n", "Epoch 3/10\n", - "325/325 [==============================] - 0s 64us/sample - loss: 1.3338 - accuracy: 0.6400\n", + "342/342 [==============================] - 0s 60us/sample - loss: 1.4207 - accuracy: 0.5614\n", "Epoch 4/10\n", - "325/325 [==============================] - 0s 63us/sample - loss: 0.9804 - accuracy: 0.7323\n", + "342/342 [==============================] - 0s 66us/sample - loss: 1.1300 - accuracy: 0.6404\n", "Epoch 5/10\n", - "325/325 [==============================] - 0s 65us/sample - loss: 0.8425 - accuracy: 0.7108\n", + "342/342 [==============================] - 0s 59us/sample - loss: 0.8926 - accuracy: 0.7193\n", "Epoch 6/10\n", - "325/325 [==============================] - 0s 61us/sample - loss: 0.6542 - accuracy: 0.8031\n", + "342/342 [==============================] - 0s 58us/sample - loss: 0.7141 - accuracy: 0.7515\n", "Epoch 7/10\n", - "325/325 [==============================] - 0s 62us/sample - loss: 0.5870 - accuracy: 0.8000\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.6551 - accuracy: 0.7807\n", "Epoch 8/10\n", - "325/325 [==============================] - 0s 64us/sample - loss: 0.4539 - accuracy: 0.8585\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.5618 - accuracy: 0.8099\n", "Epoch 9/10\n", - "325/325 [==============================] - 0s 64us/sample - loss: 0.4065 - accuracy: 0.8769\n", + "342/342 [==============================] - 0s 59us/sample - loss: 0.4506 - accuracy: 0.8567\n", "Epoch 10/10\n", - "325/325 [==============================] - 0s 64us/sample - loss: 0.3758 - accuracy: 0.8769\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.3670 - accuracy: 0.8801\n", "Train on 325 samples\n", "Epoch 1/10\n", - "325/325 [==============================] - 1s 2ms/sample - loss: 2.2283 - accuracy: 0.1969\n", + "325/325 [==============================] - 11s 34ms/sample - loss: 2.2454 - accuracy: 0.1569\n", "Epoch 2/10\n", - "325/325 [==============================] - 0s 63us/sample - loss: 1.7356 - accuracy: 0.5138\n", + "325/325 [==============================] - 0s 67us/sample - loss: 1.8617 - accuracy: 0.4585\n", "Epoch 3/10\n", - "325/325 [==============================] - 0s 63us/sample - loss: 1.2632 - accuracy: 0.6215\n", + "325/325 [==============================] - 0s 63us/sample - loss: 1.3664 - accuracy: 0.6154\n", "Epoch 4/10\n", - "325/325 [==============================] - 0s 60us/sample - loss: 0.9184 - accuracy: 0.6923\n", + "325/325 [==============================] - 0s 64us/sample - loss: 0.9389 - accuracy: 0.7138\n", "Epoch 5/10\n", - "325/325 [==============================] - 0s 62us/sample - loss: 0.8140 - accuracy: 0.7231\n", + "325/325 [==============================] - 0s 63us/sample - loss: 0.7995 - accuracy: 0.7508\n", "Epoch 6/10\n", - "325/325 [==============================] - 0s 61us/sample - loss: 0.6368 - accuracy: 0.7723\n", + "325/325 [==============================] - 0s 62us/sample - loss: 0.7195 - accuracy: 0.7600\n", "Epoch 7/10\n", - "325/325 [==============================] - 0s 62us/sample - loss: 0.5723 - accuracy: 0.7908\n", + "325/325 [==============================] - 0s 63us/sample - loss: 0.5517 - accuracy: 0.8554\n", "Epoch 8/10\n", - "325/325 [==============================] - 0s 69us/sample - loss: 0.4466 - accuracy: 0.8462\n", + "325/325 [==============================] - 0s 60us/sample - loss: 0.4940 - accuracy: 0.8585\n", "Epoch 9/10\n", - "325/325 [==============================] - 0s 60us/sample - loss: 0.4120 - accuracy: 0.8708\n", + "325/325 [==============================] - 0s 62us/sample - loss: 0.4365 - accuracy: 0.8800\n", "Epoch 10/10\n", - "325/325 [==============================] - ETA: 0s - loss: 0.3639 - accuracy: 0.89 - 0s 62us/sample - loss: 0.3645 - accuracy: 0.8862\n", - "Train on 327 samples\n", + "325/325 [==============================] - 0s 64us/sample - loss: 0.3165 - accuracy: 0.9077\n", + "Train on 322 samples\n", "Epoch 1/10\n", - "327/327 [==============================] - 11s 33ms/sample - loss: 2.2042 - accuracy: 0.2691\n", + "322/322 [==============================] - 11s 34ms/sample - loss: 2.2278 - accuracy: 0.1894\n", "Epoch 2/10\n", - "327/327 [==============================] - 0s 64us/sample - loss: 1.7250 - accuracy: 0.4954\n", + "322/322 [==============================] - 0s 69us/sample - loss: 1.7858 - accuracy: 0.5155\n", "Epoch 3/10\n", - "327/327 [==============================] - 0s 62us/sample - loss: 1.2587 - accuracy: 0.6300\n", + "322/322 [==============================] - 0s 65us/sample - loss: 1.2739 - accuracy: 0.6522\n", "Epoch 4/10\n", - "327/327 [==============================] - 0s 60us/sample - loss: 0.8851 - accuracy: 0.7064\n", + "322/322 [==============================] - 0s 64us/sample - loss: 0.9472 - accuracy: 0.6925\n", "Epoch 5/10\n", - "327/327 [==============================] - 0s 61us/sample - loss: 0.7274 - accuracy: 0.7676\n", + "322/322 [==============================] - 0s 62us/sample - loss: 0.8731 - accuracy: 0.7422\n", "Epoch 6/10\n", - "327/327 [==============================] - 0s 61us/sample - loss: 0.6097 - accuracy: 0.7951\n", + "322/322 [==============================] - 0s 61us/sample - loss: 0.6539 - accuracy: 0.7919\n", "Epoch 7/10\n", - "327/327 [==============================] - 0s 60us/sample - loss: 0.5112 - accuracy: 0.8196\n", + "322/322 [==============================] - 0s 66us/sample - loss: 0.5807 - accuracy: 0.8168\n", "Epoch 8/10\n", - "327/327 [==============================] - 0s 60us/sample - loss: 0.4093 - accuracy: 0.8869\n", + "322/322 [==============================] - 0s 63us/sample - loss: 0.4801 - accuracy: 0.8447\n", "Epoch 9/10\n", - "327/327 [==============================] - 0s 60us/sample - loss: 0.3015 - accuracy: 0.9083\n", + "322/322 [==============================] - 0s 66us/sample - loss: 0.4348 - accuracy: 0.8602\n", "Epoch 10/10\n", - "327/327 [==============================] - 0s 58us/sample - loss: 0.3232 - accuracy: 0.9021\n", - "Train on 354 samples\n", + "322/322 [==============================] - 0s 62us/sample - loss: 0.2882 - accuracy: 0.9161\n", + "Train on 347 samples\n", "Epoch 1/10\n", - "354/354 [==============================] - 11s 31ms/sample - loss: 2.2384 - accuracy: 0.1610\n", + "347/347 [==============================] - 11s 32ms/sample - loss: 2.2546 - accuracy: 0.1210\n", "Epoch 2/10\n", - "354/354 [==============================] - 0s 62us/sample - loss: 1.7841 - accuracy: 0.4718\n", + "347/347 [==============================] - 0s 64us/sample - loss: 1.8273 - accuracy: 0.4841\n", "Epoch 3/10\n", - "354/354 [==============================] - 0s 58us/sample - loss: 1.3495 - accuracy: 0.6045\n", + "347/347 [==============================] - 0s 60us/sample - loss: 1.4393 - accuracy: 0.5620\n", "Epoch 4/10\n", - "354/354 [==============================] - 0s 58us/sample - loss: 1.0376 - accuracy: 0.6638\n", + "347/347 [==============================] - 0s 63us/sample - loss: 1.0664 - accuracy: 0.6571\n", "Epoch 5/10\n", - "354/354 [==============================] - 0s 60us/sample - loss: 0.8677 - accuracy: 0.7090\n", + "347/347 [==============================] - 0s 60us/sample - loss: 0.9386 - accuracy: 0.6916\n", "Epoch 6/10\n", - "354/354 [==============================] - 0s 57us/sample - loss: 0.6407 - accuracy: 0.7712\n", + "347/347 [==============================] - 0s 59us/sample - loss: 0.7910 - accuracy: 0.7493\n", "Epoch 7/10\n", - "354/354 [==============================] - 0s 59us/sample - loss: 0.6079 - accuracy: 0.7966\n", + "347/347 [==============================] - 0s 61us/sample - loss: 0.6547 - accuracy: 0.7983\n", "Epoch 8/10\n", - "354/354 [==============================] - 0s 80us/sample - loss: 0.4989 - accuracy: 0.8531\n", + "347/347 [==============================] - 0s 69us/sample - loss: 0.5288 - accuracy: 0.8559\n", "Epoch 9/10\n", - "354/354 [==============================] - 0s 58us/sample - loss: 0.4501 - accuracy: 0.8588\n", + "347/347 [==============================] - 0s 63us/sample - loss: 0.5278 - accuracy: 0.8127\n", "Epoch 10/10\n", - "354/354 [==============================] - 0s 58us/sample - loss: 0.3984 - accuracy: 0.8701\n", - "Train on 317 samples\n", + "347/347 [==============================] - 0s 60us/sample - loss: 0.3877 - accuracy: 0.8761\n", + "Train on 309 samples\n", "Epoch 1/10\n", - "317/317 [==============================] - 11s 35ms/sample - loss: 2.2234 - accuracy: 0.2334\n", + "309/309 [==============================] - 11s 36ms/sample - loss: 2.2360 - accuracy: 0.2071\n", "Epoch 2/10\n", - "317/317 [==============================] - 0s 67us/sample - loss: 1.7150 - accuracy: 0.5426\n", + "309/309 [==============================] - 0s 70us/sample - loss: 1.7816 - accuracy: 0.5243\n", "Epoch 3/10\n", - "317/317 [==============================] - 0s 63us/sample - loss: 1.1822 - accuracy: 0.6593\n", + "309/309 [==============================] - 0s 65us/sample - loss: 1.2704 - accuracy: 0.6667\n", "Epoch 4/10\n", - "317/317 [==============================] - 0s 65us/sample - loss: 0.8759 - accuracy: 0.7161\n", + "309/309 [==============================] - 0s 64us/sample - loss: 0.9334 - accuracy: 0.6990\n", "Epoch 5/10\n", - "317/317 [==============================] - 0s 66us/sample - loss: 0.6869 - accuracy: 0.7697\n", + "309/309 [==============================] - 0s 63us/sample - loss: 0.7528 - accuracy: 0.7443\n", "Epoch 6/10\n", - "317/317 [==============================] - 0s 63us/sample - loss: 0.5508 - accuracy: 0.8013\n", + "309/309 [==============================] - 0s 64us/sample - loss: 0.6303 - accuracy: 0.7929\n", "Epoch 7/10\n", - "317/317 [==============================] - 0s 63us/sample - loss: 0.4736 - accuracy: 0.8360\n", + "309/309 [==============================] - 0s 63us/sample - loss: 0.5035 - accuracy: 0.8220\n", "Epoch 8/10\n", - "317/317 [==============================] - 0s 60us/sample - loss: 0.4492 - accuracy: 0.8612\n", + "309/309 [==============================] - 0s 78us/sample - loss: 0.4660 - accuracy: 0.8479\n", "Epoch 9/10\n", - "317/317 [==============================] - 0s 61us/sample - loss: 0.4136 - accuracy: 0.8927\n", + "309/309 [==============================] - 0s 70us/sample - loss: 0.4100 - accuracy: 0.8511\n", "Epoch 10/10\n", - "317/317 [==============================] - 0s 68us/sample - loss: 0.3069 - accuracy: 0.9117\n", - "Train on 321 samples\n", + "309/309 [==============================] - 0s 73us/sample - loss: 0.3488 - accuracy: 0.8867\n", + "Train on 308 samples\n", "Epoch 1/10\n", - "321/321 [==============================] - 1s 2ms/sample - loss: 2.2172 - accuracy: 0.1620\n", + "308/308 [==============================] - 11s 36ms/sample - loss: 2.2350 - accuracy: 0.1786\n", "Epoch 2/10\n", - "321/321 [==============================] - 0s 88us/sample - loss: 1.8149 - accuracy: 0.4174\n", + "308/308 [==============================] - 0s 70us/sample - loss: 1.8466 - accuracy: 0.4351\n", "Epoch 3/10\n", - "321/321 [==============================] - 0s 62us/sample - loss: 1.3609 - accuracy: 0.5794\n", + "308/308 [==============================] - 0s 64us/sample - loss: 1.3188 - accuracy: 0.6169\n", "Epoch 4/10\n", - "321/321 [==============================] - 0s 62us/sample - loss: 1.0089 - accuracy: 0.6760\n", + "308/308 [==============================] - 0s 64us/sample - loss: 1.0252 - accuracy: 0.6526\n", "Epoch 5/10\n", - "321/321 [==============================] - 0s 61us/sample - loss: 0.7772 - accuracy: 0.7259\n", + "308/308 [==============================] - 0s 65us/sample - loss: 0.7959 - accuracy: 0.7370\n", "Epoch 6/10\n" ] }, @@ -1429,167 +1457,167 @@ "name": "stdout", "output_type": "stream", "text": [ - "321/321 [==============================] - 0s 61us/sample - loss: 0.6457 - accuracy: 0.7819\n", + "308/308 [==============================] - 0s 63us/sample - loss: 0.7043 - accuracy: 0.7500\n", "Epoch 7/10\n", - "321/321 [==============================] - 0s 79us/sample - loss: 0.5311 - accuracy: 0.8255\n", + "308/308 [==============================] - 0s 64us/sample - loss: 0.5364 - accuracy: 0.8247\n", "Epoch 8/10\n", - "321/321 [==============================] - 0s 64us/sample - loss: 0.4651 - accuracy: 0.8505\n", + "308/308 [==============================] - 0s 64us/sample - loss: 0.4573 - accuracy: 0.8539\n", "Epoch 9/10\n", - "321/321 [==============================] - 0s 60us/sample - loss: 0.4384 - accuracy: 0.8442\n", + "308/308 [==============================] - 0s 65us/sample - loss: 0.4340 - accuracy: 0.8506\n", "Epoch 10/10\n", - "321/321 [==============================] - 0s 60us/sample - loss: 0.3757 - accuracy: 0.8660\n", - "Train on 337 samples\n", + "308/308 [==============================] - 0s 64us/sample - loss: 0.3776 - accuracy: 0.8734\n", + "Train on 335 samples\n", "Epoch 1/10\n", - "337/337 [==============================] - 11s 33ms/sample - loss: 2.2155 - accuracy: 0.2226\n", + "335/335 [==============================] - 11s 34ms/sample - loss: 2.2041 - accuracy: 0.1881\n", "Epoch 2/10\n", - "337/337 [==============================] - 0s 66us/sample - loss: 1.7212 - accuracy: 0.4866\n", + "335/335 [==============================] - 0s 69us/sample - loss: 1.7178 - accuracy: 0.5134\n", "Epoch 3/10\n", - "337/337 [==============================] - 0s 71us/sample - loss: 1.2952 - accuracy: 0.5727\n", + "335/335 [==============================] - 0s 63us/sample - loss: 1.2325 - accuracy: 0.6418\n", "Epoch 4/10\n", - "337/337 [==============================] - 0s 61us/sample - loss: 0.9156 - accuracy: 0.6736\n", + "335/335 [==============================] - 0s 62us/sample - loss: 0.9610 - accuracy: 0.6955\n", "Epoch 5/10\n", - "337/337 [==============================] - 0s 61us/sample - loss: 0.7012 - accuracy: 0.7537\n", + "335/335 [==============================] - 0s 62us/sample - loss: 0.7274 - accuracy: 0.7433\n", "Epoch 6/10\n", - "337/337 [==============================] - 0s 62us/sample - loss: 0.6495 - accuracy: 0.8131\n", + "335/335 [==============================] - 0s 62us/sample - loss: 0.5745 - accuracy: 0.8030\n", "Epoch 7/10\n", - "337/337 [==============================] - 0s 61us/sample - loss: 0.4682 - accuracy: 0.8576\n", + "335/335 [==============================] - 0s 61us/sample - loss: 0.5069 - accuracy: 0.8507\n", "Epoch 8/10\n", - "337/337 [==============================] - 0s 65us/sample - loss: 0.4705 - accuracy: 0.8398\n", + "335/335 [==============================] - 0s 76us/sample - loss: 0.4238 - accuracy: 0.8687\n", "Epoch 9/10\n", - "337/337 [==============================] - 0s 63us/sample - loss: 0.3500 - accuracy: 0.9110\n", + "335/335 [==============================] - 0s 79us/sample - loss: 0.3895 - accuracy: 0.8836\n", "Epoch 10/10\n", - "337/337 [==============================] - 0s 61us/sample - loss: 0.3329 - accuracy: 0.9050\n", - "Train on 307 samples\n", + "335/335 [==============================] - 0s 62us/sample - loss: 0.3400 - accuracy: 0.8925\n", + "Train on 302 samples\n", "Epoch 1/10\n", - "307/307 [==============================] - 11s 37ms/sample - loss: 2.2308 - accuracy: 0.1922\n", + "302/302 [==============================] - 11s 38ms/sample - loss: 2.2356 - accuracy: 0.1788\n", "Epoch 2/10\n", - "307/307 [==============================] - 0s 66us/sample - loss: 1.8578 - accuracy: 0.4235\n", + "302/302 [==============================] - 0s 70us/sample - loss: 1.8192 - accuracy: 0.4735\n", "Epoch 3/10\n", - "307/307 [==============================] - 0s 65us/sample - loss: 1.2925 - accuracy: 0.6384\n", + "302/302 [==============================] - 0s 64us/sample - loss: 1.3145 - accuracy: 0.6325\n", "Epoch 4/10\n", - "307/307 [==============================] - 0s 63us/sample - loss: 1.0488 - accuracy: 0.6612\n", + "302/302 [==============================] - 0s 62us/sample - loss: 0.9660 - accuracy: 0.6854\n", "Epoch 5/10\n", - "307/307 [==============================] - 0s 62us/sample - loss: 0.7850 - accuracy: 0.7264\n", + "302/302 [==============================] - 0s 63us/sample - loss: 0.8150 - accuracy: 0.7351\n", "Epoch 6/10\n", - "307/307 [==============================] - 0s 64us/sample - loss: 0.6545 - accuracy: 0.8046\n", + "302/302 [==============================] - 0s 63us/sample - loss: 0.6891 - accuracy: 0.7781\n", "Epoch 7/10\n", - "307/307 [==============================] - 0s 61us/sample - loss: 0.5507 - accuracy: 0.8339\n", + "302/302 [==============================] - 0s 64us/sample - loss: 0.5861 - accuracy: 0.8179\n", "Epoch 8/10\n", - "307/307 [==============================] - 0s 73us/sample - loss: 0.4836 - accuracy: 0.8502\n", + "302/302 [==============================] - 0s 63us/sample - loss: 0.5031 - accuracy: 0.8510\n", "Epoch 9/10\n", - "307/307 [==============================] - 0s 63us/sample - loss: 0.4530 - accuracy: 0.8664\n", + "302/302 [==============================] - 0s 68us/sample - loss: 0.4585 - accuracy: 0.8675\n", "Epoch 10/10\n", - "307/307 [==============================] - 0s 62us/sample - loss: 0.3767 - accuracy: 0.8860\n", - "Train on 341 samples\n", + "302/302 [==============================] - 0s 64us/sample - loss: 0.4087 - accuracy: 0.8675\n", + "Train on 315 samples\n", "Epoch 1/10\n", - "341/341 [==============================] - 11s 33ms/sample - loss: 2.2294 - accuracy: 0.2141\n", + "315/315 [==============================] - 12s 37ms/sample - loss: 2.2341 - accuracy: 0.1810\n", "Epoch 2/10\n", - "341/341 [==============================] - 0s 65us/sample - loss: 1.7668 - accuracy: 0.4809\n", + "315/315 [==============================] - 0s 80us/sample - loss: 1.8170 - accuracy: 0.4159\n", "Epoch 3/10\n", - "341/341 [==============================] - 0s 61us/sample - loss: 1.2499 - accuracy: 0.6481\n", + "315/315 [==============================] - 0s 66us/sample - loss: 1.3814 - accuracy: 0.6095\n", "Epoch 4/10\n", - "341/341 [==============================] - 0s 61us/sample - loss: 0.8826 - accuracy: 0.7185\n", + "315/315 [==============================] - 0s 64us/sample - loss: 1.0259 - accuracy: 0.6730\n", "Epoch 5/10\n", - "341/341 [==============================] - 0s 60us/sample - loss: 0.6781 - accuracy: 0.8152\n", + "315/315 [==============================] - 0s 66us/sample - loss: 0.9151 - accuracy: 0.6825\n", "Epoch 6/10\n", - "341/341 [==============================] - 0s 65us/sample - loss: 0.5634 - accuracy: 0.8065\n", + "315/315 [==============================] - 0s 66us/sample - loss: 0.6886 - accuracy: 0.7524\n", "Epoch 7/10\n", - "341/341 [==============================] - 0s 60us/sample - loss: 0.5126 - accuracy: 0.8270\n", + "315/315 [==============================] - 0s 63us/sample - loss: 0.6590 - accuracy: 0.7841\n", "Epoch 8/10\n", - "341/341 [==============================] - 0s 57us/sample - loss: 0.4333 - accuracy: 0.8416\n", + "315/315 [==============================] - 0s 67us/sample - loss: 0.4832 - accuracy: 0.8635\n", "Epoch 9/10\n", - "341/341 [==============================] - 0s 59us/sample - loss: 0.3950 - accuracy: 0.8710\n", + "315/315 [==============================] - 0s 66us/sample - loss: 0.4498 - accuracy: 0.8635\n", "Epoch 10/10\n", - "341/341 [==============================] - 0s 59us/sample - loss: 0.3252 - accuracy: 0.8944\n", - "Train on 366 samples\n", + "315/315 [==============================] - 0s 64us/sample - loss: 0.3939 - accuracy: 0.8762\n", + "Train on 345 samples\n", "Epoch 1/10\n", - "366/366 [==============================] - 11s 31ms/sample - loss: 2.2324 - accuracy: 0.2186\n", + "345/345 [==============================] - 12s 33ms/sample - loss: 2.2156 - accuracy: 0.1942\n", "Epoch 2/10\n", - "366/366 [==============================] - 0s 64us/sample - loss: 1.7776 - accuracy: 0.5137\n", + "345/345 [==============================] - 0s 71us/sample - loss: 1.7534 - accuracy: 0.5275\n", "Epoch 3/10\n", - "366/366 [==============================] - 0s 60us/sample - loss: 1.2962 - accuracy: 0.6257\n", + "345/345 [==============================] - 0s 66us/sample - loss: 1.2267 - accuracy: 0.6609\n", "Epoch 4/10\n", - "366/366 [==============================] - 0s 59us/sample - loss: 0.9391 - accuracy: 0.7240\n", + "345/345 [==============================] - 0s 62us/sample - loss: 0.9566 - accuracy: 0.6928\n", "Epoch 5/10\n", - "366/366 [==============================] - 0s 59us/sample - loss: 0.7795 - accuracy: 0.7514\n", + "345/345 [==============================] - 0s 63us/sample - loss: 0.7458 - accuracy: 0.7942\n", "Epoch 6/10\n", - "366/366 [==============================] - 0s 59us/sample - loss: 0.6495 - accuracy: 0.7869\n", + "345/345 [==============================] - 0s 72us/sample - loss: 0.6715 - accuracy: 0.7884\n", "Epoch 7/10\n", - "366/366 [==============================] - 0s 58us/sample - loss: 0.5224 - accuracy: 0.8251\n", + "345/345 [==============================] - 0s 60us/sample - loss: 0.5645 - accuracy: 0.8261\n", "Epoch 8/10\n", - "366/366 [==============================] - 0s 58us/sample - loss: 0.4513 - accuracy: 0.8634\n", + "345/345 [==============================] - 0s 60us/sample - loss: 0.4883 - accuracy: 0.8464\n", "Epoch 9/10\n", - "366/366 [==============================] - 0s 58us/sample - loss: 0.3704 - accuracy: 0.8934\n", + "345/345 [==============================] - 0s 63us/sample - loss: 0.4045 - accuracy: 0.8696\n", "Epoch 10/10\n", - "366/366 [==============================] - 0s 57us/sample - loss: 0.2898 - accuracy: 0.9317\n", - "Train on 323 samples\n", + "345/345 [==============================] - 0s 62us/sample - loss: 0.3806 - accuracy: 0.8812\n", + "Train on 360 samples\n", "Epoch 1/10\n", - "323/323 [==============================] - 11s 35ms/sample - loss: 2.1944 - accuracy: 0.2817\n", + "360/360 [==============================] - 1s 2ms/sample - loss: 2.1870 - accuracy: 0.2000\n", "Epoch 2/10\n", - "323/323 [==============================] - 0s 68us/sample - loss: 1.7057 - accuracy: 0.5201\n", + "360/360 [==============================] - 0s 60us/sample - loss: 1.6846 - accuracy: 0.5583\n", "Epoch 3/10\n", - "323/323 [==============================] - 0s 62us/sample - loss: 1.1160 - accuracy: 0.6811\n", + "360/360 [==============================] - 0s 62us/sample - loss: 1.2173 - accuracy: 0.6028\n", "Epoch 4/10\n", - "323/323 [==============================] - 0s 61us/sample - loss: 0.8963 - accuracy: 0.6935\n", + "360/360 [==============================] - 0s 59us/sample - loss: 0.9515 - accuracy: 0.7028\n", "Epoch 5/10\n", - "323/323 [==============================] - 0s 61us/sample - loss: 0.7364 - accuracy: 0.7554\n", + "360/360 [==============================] - 0s 60us/sample - loss: 0.7086 - accuracy: 0.7806\n", "Epoch 6/10\n", - "323/323 [==============================] - 0s 61us/sample - loss: 0.6568 - accuracy: 0.7771\n", + "360/360 [==============================] - 0s 63us/sample - loss: 0.6152 - accuracy: 0.8056\n", "Epoch 7/10\n", - "323/323 [==============================] - 0s 61us/sample - loss: 0.5393 - accuracy: 0.8359\n", + "360/360 [==============================] - 0s 64us/sample - loss: 0.5507 - accuracy: 0.8139\n", "Epoch 8/10\n", - "323/323 [==============================] - 0s 61us/sample - loss: 0.4105 - accuracy: 0.8793\n", + "360/360 [==============================] - 0s 61us/sample - loss: 0.4329 - accuracy: 0.8611\n", "Epoch 9/10\n", - "323/323 [==============================] - 0s 62us/sample - loss: 0.3923 - accuracy: 0.8824\n", + "360/360 [==============================] - 0s 65us/sample - loss: 0.3778 - accuracy: 0.8944\n", "Epoch 10/10\n", - "323/323 [==============================] - 0s 61us/sample - loss: 0.3041 - accuracy: 0.9133\n", - "Train on 314 samples\n", + "360/360 [==============================] - 0s 60us/sample - loss: 0.3881 - accuracy: 0.8778\n", + "Train on 343 samples\n", "Epoch 1/10\n", - "314/314 [==============================] - 11s 36ms/sample - loss: 2.2331 - accuracy: 0.2038\n", + "343/343 [==============================] - 12s 35ms/sample - loss: 2.2056 - accuracy: 0.2099\n", "Epoch 2/10\n", - "314/314 [==============================] - 0s 70us/sample - loss: 1.7858 - accuracy: 0.5223\n", + "343/343 [==============================] - 0s 67us/sample - loss: 1.7478 - accuracy: 0.4519\n", "Epoch 3/10\n", - "314/314 [==============================] - 0s 65us/sample - loss: 1.2197 - accuracy: 0.6656\n", + "343/343 [==============================] - 0s 64us/sample - loss: 1.2228 - accuracy: 0.6297\n", "Epoch 4/10\n", - "314/314 [==============================] - 0s 62us/sample - loss: 0.9685 - accuracy: 0.6911\n", + "343/343 [==============================] - 0s 62us/sample - loss: 0.9265 - accuracy: 0.7230\n", "Epoch 5/10\n", - "314/314 [==============================] - 0s 64us/sample - loss: 0.8166 - accuracy: 0.7325\n", + "343/343 [==============================] - 0s 62us/sample - loss: 0.7601 - accuracy: 0.7493\n", "Epoch 6/10\n", - "314/314 [==============================] - 0s 64us/sample - loss: 0.6338 - accuracy: 0.7930\n", + "343/343 [==============================] - 0s 63us/sample - loss: 0.5683 - accuracy: 0.8309\n", "Epoch 7/10\n", - "314/314 [==============================] - 0s 64us/sample - loss: 0.5093 - accuracy: 0.8599\n", + "343/343 [==============================] - 0s 64us/sample - loss: 0.4596 - accuracy: 0.8513\n", "Epoch 8/10\n", - "314/314 [==============================] - 0s 65us/sample - loss: 0.3852 - accuracy: 0.9013\n", + "343/343 [==============================] - 0s 63us/sample - loss: 0.3302 - accuracy: 0.8921\n", "Epoch 9/10\n", - "314/314 [==============================] - 0s 66us/sample - loss: 0.3595 - accuracy: 0.8885\n", + "343/343 [==============================] - 0s 61us/sample - loss: 0.2962 - accuracy: 0.9067\n", "Epoch 10/10\n", - "314/314 [==============================] - 0s 65us/sample - loss: 0.3047 - accuracy: 0.9045\n", - "Train on 334 samples\n", + "343/343 [==============================] - 0s 61us/sample - loss: 0.2488 - accuracy: 0.9184\n", + "Train on 342 samples\n", "Epoch 1/10\n", - "334/334 [==============================] - 12s 35ms/sample - loss: 2.2275 - accuracy: 0.2156\n", + "342/342 [==============================] - 1s 2ms/sample - loss: 2.2205 - accuracy: 0.2398\n", "Epoch 2/10\n", - "334/334 [==============================] - 0s 67us/sample - loss: 1.7395 - accuracy: 0.5329\n", + "342/342 [==============================] - 0s 62us/sample - loss: 1.7411 - accuracy: 0.5673\n", "Epoch 3/10\n", - "334/334 [==============================] - 0s 64us/sample - loss: 1.2879 - accuracy: 0.6407\n", + "342/342 [==============================] - 0s 61us/sample - loss: 1.2467 - accuracy: 0.6404\n", "Epoch 4/10\n", - "334/334 [==============================] - 0s 63us/sample - loss: 0.9852 - accuracy: 0.6856\n", + "342/342 [==============================] - 0s 63us/sample - loss: 0.8607 - accuracy: 0.7222\n", "Epoch 5/10\n", - "334/334 [==============================] - 0s 61us/sample - loss: 0.7367 - accuracy: 0.7575\n", + "342/342 [==============================] - 0s 62us/sample - loss: 0.6752 - accuracy: 0.7865\n", "Epoch 6/10\n", - "334/334 [==============================] - 0s 62us/sample - loss: 0.5724 - accuracy: 0.8204\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.6142 - accuracy: 0.8070\n", "Epoch 7/10\n", - "334/334 [==============================] - 0s 63us/sample - loss: 0.4697 - accuracy: 0.8533\n", + "342/342 [==============================] - 0s 64us/sample - loss: 0.4820 - accuracy: 0.8509\n", "Epoch 8/10\n", - "334/334 [==============================] - 0s 62us/sample - loss: 0.3997 - accuracy: 0.8683\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.4223 - accuracy: 0.8713\n", "Epoch 9/10\n", - "334/334 [==============================] - 0s 60us/sample - loss: 0.3321 - accuracy: 0.9012\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.3845 - accuracy: 0.8801\n", "Epoch 10/10\n", - "334/334 [==============================] - 0s 70us/sample - loss: 0.3126 - accuracy: 0.8713\n", - "Train on 329 samples\n", + "342/342 [==============================] - 0s 61us/sample - loss: 0.3521 - accuracy: 0.8567\n", + "Train on 330 samples\n", "Epoch 1/10\n", - "329/329 [==============================] - 12s 35ms/sample - loss: 2.2420 - accuracy: 0.1793\n", + "330/330 [==============================] - 12s 36ms/sample - loss: 2.2459 - accuracy: 0.2000\n", "Epoch 2/10\n", - "329/329 [==============================] - 0s 66us/sample - loss: 1.8302 - accuracy: 0.4134\n", + "330/330 [==============================] - 0s 68us/sample - loss: 1.8403 - accuracy: 0.5212\n", "Epoch 3/10\n" ] }, @@ -1597,169 +1625,169 @@ "name": "stdout", "output_type": "stream", "text": [ - "329/329 [==============================] - 0s 73us/sample - loss: 1.3702 - accuracy: 0.6292\n", + "330/330 [==============================] - 0s 62us/sample - loss: 1.3075 - accuracy: 0.6303\n", "Epoch 4/10\n", - "329/329 [==============================] - 0s 63us/sample - loss: 1.0237 - accuracy: 0.6748\n", + "330/330 [==============================] - 0s 63us/sample - loss: 0.9618 - accuracy: 0.7091\n", "Epoch 5/10\n", - "329/329 [==============================] - 0s 63us/sample - loss: 0.7861 - accuracy: 0.7690\n", + "330/330 [==============================] - 0s 63us/sample - loss: 0.7600 - accuracy: 0.7576\n", "Epoch 6/10\n", - "329/329 [==============================] - 0s 61us/sample - loss: 0.6595 - accuracy: 0.7933\n", + "330/330 [==============================] - 0s 64us/sample - loss: 0.6339 - accuracy: 0.7848\n", "Epoch 7/10\n", - "329/329 [==============================] - 0s 61us/sample - loss: 0.5333 - accuracy: 0.8055\n", + "330/330 [==============================] - 0s 65us/sample - loss: 0.5387 - accuracy: 0.8182\n", "Epoch 8/10\n", - "329/329 [==============================] - 0s 61us/sample - loss: 0.4790 - accuracy: 0.8480\n", + "330/330 [==============================] - 0s 62us/sample - loss: 0.4587 - accuracy: 0.8576\n", "Epoch 9/10\n", - "329/329 [==============================] - 0s 62us/sample - loss: 0.4028 - accuracy: 0.8754\n", + "330/330 [==============================] - 0s 61us/sample - loss: 0.3809 - accuracy: 0.8848\n", "Epoch 10/10\n", - "329/329 [==============================] - 0s 62us/sample - loss: 0.3721 - accuracy: 0.8815\n", - "Train on 330 samples\n", + "330/330 [==============================] - 0s 67us/sample - loss: 0.3699 - accuracy: 0.8697\n", + "Train on 318 samples\n", "Epoch 1/10\n", - "330/330 [==============================] - 12s 35ms/sample - loss: 2.2205 - accuracy: 0.2061\n", + "318/318 [==============================] - 12s 37ms/sample - loss: 2.1802 - accuracy: 0.2013\n", "Epoch 2/10\n", - "330/330 [==============================] - 0s 68us/sample - loss: 1.7812 - accuracy: 0.4576\n", + "318/318 [==============================] - 0s 71us/sample - loss: 1.7196 - accuracy: 0.5000\n", "Epoch 3/10\n", - "330/330 [==============================] - 0s 62us/sample - loss: 1.2546 - accuracy: 0.6424\n", + "318/318 [==============================] - 0s 66us/sample - loss: 1.1923 - accuracy: 0.6572\n", "Epoch 4/10\n", - "330/330 [==============================] - 0s 62us/sample - loss: 0.9477 - accuracy: 0.6970\n", + "318/318 [==============================] - 0s 64us/sample - loss: 0.8613 - accuracy: 0.6950\n", "Epoch 5/10\n", - "330/330 [==============================] - 0s 61us/sample - loss: 0.7458 - accuracy: 0.7576\n", + "318/318 [==============================] - 0s 66us/sample - loss: 0.7403 - accuracy: 0.7390\n", "Epoch 6/10\n", - "330/330 [==============================] - 0s 60us/sample - loss: 0.5802 - accuracy: 0.7939\n", + "318/318 [==============================] - 0s 65us/sample - loss: 0.5980 - accuracy: 0.7862\n", "Epoch 7/10\n", - "330/330 [==============================] - 0s 62us/sample - loss: 0.5273 - accuracy: 0.8545\n", + "318/318 [==============================] - 0s 66us/sample - loss: 0.4876 - accuracy: 0.8333\n", "Epoch 8/10\n", - "330/330 [==============================] - 0s 63us/sample - loss: 0.4467 - accuracy: 0.8606\n", + "318/318 [==============================] - 0s 84us/sample - loss: 0.4420 - accuracy: 0.8459\n", "Epoch 9/10\n", - "330/330 [==============================] - 0s 61us/sample - loss: 0.3525 - accuracy: 0.8818\n", + "318/318 [==============================] - 0s 64us/sample - loss: 0.3826 - accuracy: 0.8648\n", "Epoch 10/10\n", - "330/330 [==============================] - 0s 63us/sample - loss: 0.2829 - accuracy: 0.9212\n", - "Train on 361 samples\n", + "318/318 [==============================] - 0s 65us/sample - loss: 0.3374 - accuracy: 0.8931\n", + "Train on 344 samples\n", "Epoch 1/10\n", - "361/361 [==============================] - 1s 2ms/sample - loss: 2.1974 - accuracy: 0.2382\n", + "344/344 [==============================] - 12s 35ms/sample - loss: 2.2292 - accuracy: 0.1570\n", "Epoch 2/10\n", - "361/361 [==============================] - 0s 68us/sample - loss: 1.7910 - accuracy: 0.4903\n", + "344/344 [==============================] - 0s 68us/sample - loss: 1.7692 - accuracy: 0.5174\n", "Epoch 3/10\n", - "361/361 [==============================] - 0s 59us/sample - loss: 1.2943 - accuracy: 0.5762\n", + "344/344 [==============================] - 0s 62us/sample - loss: 1.2617 - accuracy: 0.6308\n", "Epoch 4/10\n", - "361/361 [==============================] - 0s 60us/sample - loss: 0.9222 - accuracy: 0.6870\n", + "344/344 [==============================] - 0s 62us/sample - loss: 0.9609 - accuracy: 0.7151\n", "Epoch 5/10\n", - "361/361 [==============================] - 0s 60us/sample - loss: 0.7048 - accuracy: 0.7645\n", + "344/344 [==============================] - 0s 63us/sample - loss: 0.7661 - accuracy: 0.7471\n", "Epoch 6/10\n", - "361/361 [==============================] - 0s 60us/sample - loss: 0.6848 - accuracy: 0.7756\n", + "344/344 [==============================] - 0s 62us/sample - loss: 0.6832 - accuracy: 0.7878\n", "Epoch 7/10\n", - "361/361 [==============================] - 0s 60us/sample - loss: 0.5517 - accuracy: 0.8116\n", + "344/344 [==============================] - 0s 61us/sample - loss: 0.4812 - accuracy: 0.8430\n", "Epoch 8/10\n", - "361/361 [==============================] - 0s 62us/sample - loss: 0.4094 - accuracy: 0.8643\n", + "344/344 [==============================] - 0s 61us/sample - loss: 0.4927 - accuracy: 0.8488\n", "Epoch 9/10\n", - "361/361 [==============================] - 0s 58us/sample - loss: 0.4014 - accuracy: 0.8615\n", + "344/344 [==============================] - 0s 63us/sample - loss: 0.3984 - accuracy: 0.8808\n", "Epoch 10/10\n", - "361/361 [==============================] - 0s 59us/sample - loss: 0.2742 - accuracy: 0.8947\n", - "Train on 329 samples\n", + "344/344 [==============================] - 0s 61us/sample - loss: 0.3699 - accuracy: 0.8866\n", + "Train on 343 samples\n", "Epoch 1/10\n", - "329/329 [==============================] - 1s 2ms/sample - loss: 2.2131 - accuracy: 0.2067\n", + "343/343 [==============================] - 1s 2ms/sample - loss: 2.1873 - accuracy: 0.2420\n", "Epoch 2/10\n", - "329/329 [==============================] - 0s 61us/sample - loss: 1.7738 - accuracy: 0.4438\n", + "343/343 [==============================] - 0s 63us/sample - loss: 1.6775 - accuracy: 0.5423\n", "Epoch 3/10\n", - "329/329 [==============================] - 0s 71us/sample - loss: 1.2896 - accuracy: 0.6231\n", + "343/343 [==============================] - 0s 61us/sample - loss: 1.1820 - accuracy: 0.6356\n", "Epoch 4/10\n", - "329/329 [==============================] - 0s 62us/sample - loss: 0.9988 - accuracy: 0.6626\n", + "343/343 [==============================] - 0s 61us/sample - loss: 0.9419 - accuracy: 0.6939\n", "Epoch 5/10\n", - "329/329 [==============================] - 0s 111us/sample - loss: 0.8169 - accuracy: 0.7568\n", + "343/343 [==============================] - 0s 63us/sample - loss: 0.7003 - accuracy: 0.7638\n", "Epoch 6/10\n", - "329/329 [==============================] - 0s 60us/sample - loss: 0.6430 - accuracy: 0.8176\n", + "343/343 [==============================] - 0s 60us/sample - loss: 0.5790 - accuracy: 0.8426\n", "Epoch 7/10\n", - "329/329 [==============================] - 0s 71us/sample - loss: 0.5918 - accuracy: 0.8116\n", + "343/343 [==============================] - 0s 62us/sample - loss: 0.4988 - accuracy: 0.8717\n", "Epoch 8/10\n", - "329/329 [==============================] - 0s 61us/sample - loss: 0.4911 - accuracy: 0.8480\n", + "343/343 [==============================] - 0s 61us/sample - loss: 0.4361 - accuracy: 0.8746\n", "Epoch 9/10\n", - "329/329 [==============================] - 0s 60us/sample - loss: 0.3918 - accuracy: 0.8632\n", + "343/343 [==============================] - 0s 62us/sample - loss: 0.3998 - accuracy: 0.8863\n", "Epoch 10/10\n", - "329/329 [==============================] - 0s 65us/sample - loss: 0.3645 - accuracy: 0.8845\n", - "Train on 325 samples\n", + "343/343 [==============================] - 0s 63us/sample - loss: 0.3446 - accuracy: 0.8863\n", + "Train on 333 samples\n", "Epoch 1/10\n", - "325/325 [==============================] - 1s 2ms/sample - loss: 2.2361 - accuracy: 0.1877\n", + "333/333 [==============================] - 12s 36ms/sample - loss: 2.2107 - accuracy: 0.2012\n", "Epoch 2/10\n", - "325/325 [==============================] - 0s 62us/sample - loss: 1.7598 - accuracy: 0.5600\n", + "333/333 [==============================] - 0s 70us/sample - loss: 1.7116 - accuracy: 0.4925\n", "Epoch 3/10\n", - "325/325 [==============================] - 0s 62us/sample - loss: 1.2164 - accuracy: 0.6585\n", + "333/333 [==============================] - 0s 63us/sample - loss: 1.2445 - accuracy: 0.6096\n", "Epoch 4/10\n", - "325/325 [==============================] - 0s 70us/sample - loss: 0.8830 - accuracy: 0.7200\n", + "333/333 [==============================] - 0s 63us/sample - loss: 0.9491 - accuracy: 0.6877\n", "Epoch 5/10\n", - "325/325 [==============================] - 0s 61us/sample - loss: 0.7346 - accuracy: 0.7538\n", + "333/333 [==============================] - 0s 64us/sample - loss: 0.8191 - accuracy: 0.7297\n", "Epoch 6/10\n", - "325/325 [==============================] - 0s 60us/sample - loss: 0.5798 - accuracy: 0.8123\n", + "333/333 [==============================] - 0s 65us/sample - loss: 0.6393 - accuracy: 0.7838\n", "Epoch 7/10\n", - "325/325 [==============================] - 0s 61us/sample - loss: 0.4541 - accuracy: 0.8554\n", + "333/333 [==============================] - 0s 63us/sample - loss: 0.5971 - accuracy: 0.7868\n", "Epoch 8/10\n", - "325/325 [==============================] - 0s 62us/sample - loss: 0.3931 - accuracy: 0.8738\n", + "333/333 [==============================] - 0s 73us/sample - loss: 0.4872 - accuracy: 0.8408\n", "Epoch 9/10\n", - "325/325 [==============================] - 0s 61us/sample - loss: 0.3108 - accuracy: 0.8954\n", + "333/333 [==============================] - 0s 64us/sample - loss: 0.4835 - accuracy: 0.8559\n", "Epoch 10/10\n", - "325/325 [==============================] - 0s 61us/sample - loss: 0.3013 - accuracy: 0.8985\n", - "Train on 311 samples\n", + "333/333 [==============================] - 0s 65us/sample - loss: 0.3642 - accuracy: 0.8829\n", + "Train on 324 samples\n", "Epoch 1/10\n", - "311/311 [==============================] - 12s 38ms/sample - loss: 2.2501 - accuracy: 0.1640\n", + "324/324 [==============================] - 1s 3ms/sample - loss: 2.2290 - accuracy: 0.1698\n", "Epoch 2/10\n", - "311/311 [==============================] - 0s 68us/sample - loss: 1.8410 - accuracy: 0.5756\n", + "324/324 [==============================] - 0s 64us/sample - loss: 1.8046 - accuracy: 0.4568\n", "Epoch 3/10\n", - "311/311 [==============================] - 0s 64us/sample - loss: 1.3422 - accuracy: 0.6077\n", + "324/324 [==============================] - 0s 64us/sample - loss: 1.3126 - accuracy: 0.5957\n", "Epoch 4/10\n", - "311/311 [==============================] - 0s 65us/sample - loss: 0.8944 - accuracy: 0.7395\n", + "324/324 [==============================] - 0s 66us/sample - loss: 0.9207 - accuracy: 0.7099\n", "Epoch 5/10\n", - "311/311 [==============================] - 0s 64us/sample - loss: 0.7600 - accuracy: 0.7621\n", + "324/324 [==============================] - 0s 64us/sample - loss: 0.7506 - accuracy: 0.7531\n", "Epoch 6/10\n", - "311/311 [==============================] - 0s 63us/sample - loss: 0.6247 - accuracy: 0.8135\n", + "324/324 [==============================] - 0s 65us/sample - loss: 0.6467 - accuracy: 0.7870\n", "Epoch 7/10\n", - "311/311 [==============================] - 0s 65us/sample - loss: 0.5162 - accuracy: 0.8232\n", + "324/324 [==============================] - 0s 64us/sample - loss: 0.5215 - accuracy: 0.8210\n", "Epoch 8/10\n", - "311/311 [==============================] - 0s 64us/sample - loss: 0.3714 - accuracy: 0.8842\n", + "324/324 [==============================] - 0s 63us/sample - loss: 0.3789 - accuracy: 0.8920\n", "Epoch 9/10\n", - "311/311 [==============================] - 0s 64us/sample - loss: 0.3522 - accuracy: 0.8939\n", + "324/324 [==============================] - 0s 63us/sample - loss: 0.3973 - accuracy: 0.8642\n", "Epoch 10/10\n", - "311/311 [==============================] - 0s 64us/sample - loss: 0.3506 - accuracy: 0.9132\n", - "Train on 330 samples\n", + "324/324 [==============================] - 0s 65us/sample - loss: 0.3149 - accuracy: 0.8981\n", + "Train on 332 samples\n", "Epoch 1/10\n", - "330/330 [==============================] - 1s 2ms/sample - loss: 2.2546 - accuracy: 0.2152\n", + "332/332 [==============================] - 12s 37ms/sample - loss: 2.2283 - accuracy: 0.2018\n", "Epoch 2/10\n", - "330/330 [==============================] - 0s 62us/sample - loss: 1.8583 - accuracy: 0.4788\n", + "332/332 [==============================] - 0s 74us/sample - loss: 1.7522 - accuracy: 0.5301\n", "Epoch 3/10\n", - "330/330 [==============================] - 0s 61us/sample - loss: 1.3782 - accuracy: 0.6273\n", + "332/332 [==============================] - 0s 63us/sample - loss: 1.2310 - accuracy: 0.6386\n", "Epoch 4/10\n", - "330/330 [==============================] - 0s 61us/sample - loss: 0.9736 - accuracy: 0.6636\n", + "332/332 [==============================] - 0s 64us/sample - loss: 0.9724 - accuracy: 0.6928\n", "Epoch 5/10\n", - "330/330 [==============================] - 0s 63us/sample - loss: 0.8049 - accuracy: 0.7515\n", + "332/332 [==============================] - 0s 64us/sample - loss: 0.7570 - accuracy: 0.7440\n", "Epoch 6/10\n", - "330/330 [==============================] - 0s 60us/sample - loss: 0.6532 - accuracy: 0.7909\n", + "332/332 [==============================] - 0s 64us/sample - loss: 0.6688 - accuracy: 0.7831\n", "Epoch 7/10\n", - "330/330 [==============================] - 0s 64us/sample - loss: 0.5907 - accuracy: 0.8061\n", + "332/332 [==============================] - 0s 62us/sample - loss: 0.4865 - accuracy: 0.8554\n", "Epoch 8/10\n", - "330/330 [==============================] - 0s 60us/sample - loss: 0.5459 - accuracy: 0.8242\n", + "332/332 [==============================] - 0s 64us/sample - loss: 0.4305 - accuracy: 0.8524\n", "Epoch 9/10\n", - "330/330 [==============================] - 0s 61us/sample - loss: 0.4102 - accuracy: 0.8727\n", + "332/332 [==============================] - 0s 62us/sample - loss: 0.3408 - accuracy: 0.8916\n", "Epoch 10/10\n", - "330/330 [==============================] - 0s 67us/sample - loss: 0.4332 - accuracy: 0.8576\n", - "Train on 340 samples\n", + "332/332 [==============================] - 0s 65us/sample - loss: 0.3165 - accuracy: 0.8916\n", + "Train on 355 samples\n", "Epoch 1/10\n", - "340/340 [==============================] - 12s 36ms/sample - loss: 2.2131 - accuracy: 0.2206\n", + "355/355 [==============================] - 1s 2ms/sample - loss: 2.2140 - accuracy: 0.2366\n", "Epoch 2/10\n", - "340/340 [==============================] - 0s 66us/sample - loss: 1.7246 - accuracy: 0.5441\n", + "355/355 [==============================] - 0s 63us/sample - loss: 1.7085 - accuracy: 0.5324\n", "Epoch 3/10\n", - "340/340 [==============================] - 0s 77us/sample - loss: 1.2330 - accuracy: 0.6471\n", + "355/355 [==============================] - 0s 61us/sample - loss: 1.2116 - accuracy: 0.6563\n", "Epoch 4/10\n", - "340/340 [==============================] - 0s 61us/sample - loss: 0.8479 - accuracy: 0.7353\n", + "355/355 [==============================] - 0s 61us/sample - loss: 0.8011 - accuracy: 0.7380\n", "Epoch 5/10\n", - "340/340 [==============================] - 0s 61us/sample - loss: 0.6870 - accuracy: 0.7706\n", + "355/355 [==============================] - 0s 61us/sample - loss: 0.7563 - accuracy: 0.7380\n", "Epoch 6/10\n", - "340/340 [==============================] - 0s 62us/sample - loss: 0.5467 - accuracy: 0.8088\n", + "355/355 [==============================] - 0s 62us/sample - loss: 0.5380 - accuracy: 0.8592\n", "Epoch 7/10\n", - "340/340 [==============================] - 0s 61us/sample - loss: 0.5209 - accuracy: 0.8176\n", + "355/355 [==============================] - 0s 61us/sample - loss: 0.5049 - accuracy: 0.8366\n", "Epoch 8/10\n", - "340/340 [==============================] - 0s 60us/sample - loss: 0.3420 - accuracy: 0.8971\n", + "355/355 [==============================] - 0s 61us/sample - loss: 0.4286 - accuracy: 0.8507\n", "Epoch 9/10\n", - "340/340 [==============================] - 0s 62us/sample - loss: 0.3306 - accuracy: 0.9000\n", + "355/355 [==============================] - 0s 61us/sample - loss: 0.3375 - accuracy: 0.8986\n", "Epoch 10/10\n", - "340/340 [==============================] - 0s 63us/sample - loss: 0.2967 - accuracy: 0.9059\n", - "Train on 344 samples\n" + "355/355 [==============================] - 0s 62us/sample - loss: 0.3076 - accuracy: 0.8958\n", + "Train on 364 samples\n" ] }, { @@ -1767,151 +1795,151 @@ "output_type": "stream", "text": [ "Epoch 1/10\n", - "344/344 [==============================] - 1s 2ms/sample - loss: 2.1742 - accuracy: 0.2703\n", + "364/364 [==============================] - 1s 2ms/sample - loss: 2.1992 - accuracy: 0.2308\n", "Epoch 2/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 1.7170 - accuracy: 0.5262\n", + "364/364 [==============================] - 0s 61us/sample - loss: 1.7003 - accuracy: 0.5412\n", "Epoch 3/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 1.1922 - accuracy: 0.6279\n", + "364/364 [==============================] - 0s 64us/sample - loss: 1.2034 - accuracy: 0.6538\n", "Epoch 4/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 0.8721 - accuracy: 0.7413\n", + "364/364 [==============================] - 0s 62us/sample - loss: 0.8895 - accuracy: 0.7170\n", "Epoch 5/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.7089 - accuracy: 0.7616\n", + "364/364 [==============================] - 0s 61us/sample - loss: 0.7179 - accuracy: 0.7637\n", "Epoch 6/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 0.6076 - accuracy: 0.7936\n", + "364/364 [==============================] - 0s 62us/sample - loss: 0.5412 - accuracy: 0.8324\n", "Epoch 7/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.4682 - accuracy: 0.8517\n", + "364/364 [==============================] - 0s 62us/sample - loss: 0.4449 - accuracy: 0.8654\n", "Epoch 8/10\n", - "344/344 [==============================] - 0s 65us/sample - loss: 0.4702 - accuracy: 0.8605\n", + "364/364 [==============================] - 0s 60us/sample - loss: 0.3979 - accuracy: 0.8819\n", "Epoch 9/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.3856 - accuracy: 0.8605\n", + "364/364 [==============================] - 0s 62us/sample - loss: 0.3312 - accuracy: 0.9038\n", "Epoch 10/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.3157 - accuracy: 0.9012\n", - "Train on 348 samples\n", + "364/364 [==============================] - 0s 60us/sample - loss: 0.2683 - accuracy: 0.9148\n", + "Train on 356 samples\n", "Epoch 1/10\n", - "348/348 [==============================] - 1s 2ms/sample - loss: 2.2077 - accuracy: 0.2155\n", + "356/356 [==============================] - 12s 35ms/sample - loss: 2.1896 - accuracy: 0.2556\n", "Epoch 2/10\n", - "348/348 [==============================] - 0s 59us/sample - loss: 1.8056 - accuracy: 0.4425\n", + "356/356 [==============================] - 0s 66us/sample - loss: 1.6516 - accuracy: 0.5253\n", "Epoch 3/10\n", - "348/348 [==============================] - 0s 59us/sample - loss: 1.2433 - accuracy: 0.6580\n", + "356/356 [==============================] - 0s 61us/sample - loss: 1.2475 - accuracy: 0.6011\n", "Epoch 4/10\n", - "348/348 [==============================] - 0s 60us/sample - loss: 0.9461 - accuracy: 0.6638\n", + "356/356 [==============================] - 0s 60us/sample - loss: 0.9433 - accuracy: 0.6910\n", "Epoch 5/10\n", - "348/348 [==============================] - 0s 60us/sample - loss: 0.7645 - accuracy: 0.7672\n", + "356/356 [==============================] - 0s 62us/sample - loss: 0.7103 - accuracy: 0.7669\n", "Epoch 6/10\n", - "348/348 [==============================] - 0s 61us/sample - loss: 0.6410 - accuracy: 0.8046\n", + "356/356 [==============================] - 0s 62us/sample - loss: 0.6561 - accuracy: 0.7865\n", "Epoch 7/10\n", - "348/348 [==============================] - 0s 60us/sample - loss: 0.4984 - accuracy: 0.8477\n", + "356/356 [==============================] - 0s 60us/sample - loss: 0.5394 - accuracy: 0.8427\n", "Epoch 8/10\n", - "348/348 [==============================] - 0s 60us/sample - loss: 0.4415 - accuracy: 0.8621\n", + "356/356 [==============================] - 0s 63us/sample - loss: 0.4949 - accuracy: 0.8399\n", "Epoch 9/10\n", - "348/348 [==============================] - 0s 61us/sample - loss: 0.3583 - accuracy: 0.9023\n", + "356/356 [==============================] - 0s 63us/sample - loss: 0.3756 - accuracy: 0.8989\n", "Epoch 10/10\n", - "348/348 [==============================] - 0s 60us/sample - loss: 0.3159 - accuracy: 0.9109\n", - "Train on 344 samples\n", + "356/356 [==============================] - 0s 67us/sample - loss: 0.3581 - accuracy: 0.8904\n", + "Train on 328 samples\n", "Epoch 1/10\n", - "344/344 [==============================] - 1s 2ms/sample - loss: 2.2350 - accuracy: 0.1831\n", + "328/328 [==============================] - 1s 3ms/sample - loss: 2.2195 - accuracy: 0.2256\n", "Epoch 2/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 1.7902 - accuracy: 0.5262\n", + "328/328 [==============================] - 0s 65us/sample - loss: 1.7603 - accuracy: 0.5305\n", "Epoch 3/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 1.2940 - accuracy: 0.6250\n", + "328/328 [==============================] - 0s 72us/sample - loss: 1.3121 - accuracy: 0.5976\n", "Epoch 4/10\n", - "344/344 [==============================] - 0s 60us/sample - loss: 0.9004 - accuracy: 0.7238\n", + "328/328 [==============================] - 0s 67us/sample - loss: 0.9780 - accuracy: 0.6982\n", "Epoch 5/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.7076 - accuracy: 0.7936\n", + "328/328 [==============================] - 0s 66us/sample - loss: 0.7637 - accuracy: 0.7470\n", "Epoch 6/10\n", - "344/344 [==============================] - 0s 63us/sample - loss: 0.6113 - accuracy: 0.7849\n", + "328/328 [==============================] - 0s 65us/sample - loss: 0.5870 - accuracy: 0.8018\n", "Epoch 7/10\n", - "344/344 [==============================] - 0s 60us/sample - loss: 0.4767 - accuracy: 0.8605\n", + "328/328 [==============================] - 0s 67us/sample - loss: 0.5439 - accuracy: 0.8110\n", "Epoch 8/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.4077 - accuracy: 0.8692\n", + "328/328 [==============================] - 0s 63us/sample - loss: 0.4581 - accuracy: 0.8750\n", "Epoch 9/10\n", - "344/344 [==============================] - 0s 60us/sample - loss: 0.3878 - accuracy: 0.8750\n", + "328/328 [==============================] - 0s 64us/sample - loss: 0.4408 - accuracy: 0.8445\n", "Epoch 10/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.3123 - accuracy: 0.8983\n", - "Train on 328 samples\n", + "328/328 [==============================] - 0s 64us/sample - loss: 0.3448 - accuracy: 0.8994\n", + "Train on 341 samples\n", "Epoch 1/10\n", - "328/328 [==============================] - 12s 38ms/sample - loss: 2.2588 - accuracy: 0.1524\n", + "341/341 [==============================] - 13s 37ms/sample - loss: 2.2095 - accuracy: 0.2287\n", "Epoch 2/10\n", - "328/328 [==============================] - 0s 69us/sample - loss: 1.9413 - accuracy: 0.3872\n", + "341/341 [==============================] - 0s 77us/sample - loss: 1.7458 - accuracy: 0.5044\n", "Epoch 3/10\n", - "328/328 [==============================] - 0s 62us/sample - loss: 1.5092 - accuracy: 0.5640\n", + "341/341 [==============================] - 0s 66us/sample - loss: 1.2516 - accuracy: 0.6364\n", "Epoch 4/10\n", - "328/328 [==============================] - 0s 62us/sample - loss: 1.1548 - accuracy: 0.6311\n", + "341/341 [==============================] - 0s 64us/sample - loss: 0.9632 - accuracy: 0.6979\n", "Epoch 5/10\n", - "328/328 [==============================] - 0s 62us/sample - loss: 0.9471 - accuracy: 0.7104\n", + "341/341 [==============================] - 0s 64us/sample - loss: 0.7896 - accuracy: 0.7537\n", "Epoch 6/10\n", - "328/328 [==============================] - 0s 60us/sample - loss: 0.7079 - accuracy: 0.7652\n", + "341/341 [==============================] - 0s 67us/sample - loss: 0.5380 - accuracy: 0.8328\n", "Epoch 7/10\n", - "328/328 [==============================] - 0s 63us/sample - loss: 0.6496 - accuracy: 0.7774\n", + "341/341 [==============================] - 0s 63us/sample - loss: 0.5125 - accuracy: 0.8299\n", "Epoch 8/10\n", - "328/328 [==============================] - 0s 63us/sample - loss: 0.5642 - accuracy: 0.8110\n", + "341/341 [==============================] - 0s 66us/sample - loss: 0.4124 - accuracy: 0.8651\n", "Epoch 9/10\n", - "328/328 [==============================] - 0s 62us/sample - loss: 0.4879 - accuracy: 0.8445\n", + "341/341 [==============================] - 0s 61us/sample - loss: 0.3328 - accuracy: 0.8974\n", "Epoch 10/10\n", - "328/328 [==============================] - 0s 63us/sample - loss: 0.4408 - accuracy: 0.8659\n", - "Train on 346 samples\n", + "341/341 [==============================] - 0s 63us/sample - loss: 0.3398 - accuracy: 0.8944\n", + "Train on 334 samples\n", "Epoch 1/10\n", - "346/346 [==============================] - 12s 36ms/sample - loss: 2.2383 - accuracy: 0.1994\n", + "334/334 [==============================] - 13s 38ms/sample - loss: 2.2204 - accuracy: 0.2186\n", "Epoch 2/10\n", - "346/346 [==============================] - 0s 67us/sample - loss: 1.7799 - accuracy: 0.5549\n", + "334/334 [==============================] - 0s 70us/sample - loss: 1.7547 - accuracy: 0.4880\n", "Epoch 3/10\n", - "346/346 [==============================] - 0s 63us/sample - loss: 1.3109 - accuracy: 0.6416\n", + "334/334 [==============================] - 0s 69us/sample - loss: 1.2073 - accuracy: 0.6317\n", "Epoch 4/10\n", - "346/346 [==============================] - 0s 62us/sample - loss: 0.9413 - accuracy: 0.7081\n", + "334/334 [==============================] - 0s 68us/sample - loss: 0.9725 - accuracy: 0.6617\n", "Epoch 5/10\n", - "346/346 [==============================] - 0s 60us/sample - loss: 0.7528 - accuracy: 0.7803\n", + "334/334 [==============================] - 0s 67us/sample - loss: 0.7449 - accuracy: 0.7515\n", "Epoch 6/10\n", - "346/346 [==============================] - 0s 59us/sample - loss: 0.5846 - accuracy: 0.8121\n", + "334/334 [==============================] - 0s 66us/sample - loss: 0.6154 - accuracy: 0.7994\n", "Epoch 7/10\n", - "346/346 [==============================] - 0s 62us/sample - loss: 0.5561 - accuracy: 0.8382\n", + "334/334 [==============================] - 0s 66us/sample - loss: 0.5448 - accuracy: 0.8204\n", "Epoch 8/10\n", - "346/346 [==============================] - 0s 61us/sample - loss: 0.4421 - accuracy: 0.8728\n", + "334/334 [==============================] - 0s 66us/sample - loss: 0.4493 - accuracy: 0.8533\n", "Epoch 9/10\n", - "346/346 [==============================] - 0s 61us/sample - loss: 0.3476 - accuracy: 0.8960\n", + "334/334 [==============================] - 0s 66us/sample - loss: 0.4034 - accuracy: 0.8533\n", "Epoch 10/10\n", - "346/346 [==============================] - 0s 71us/sample - loss: 0.3717 - accuracy: 0.8931\n", - "Train on 347 samples\n", + "334/334 [==============================] - 0s 65us/sample - loss: 0.3736 - accuracy: 0.8802\n", + "Train on 313 samples\n", "Epoch 1/10\n", - "347/347 [==============================] - 12s 36ms/sample - loss: 2.2179 - accuracy: 0.2594\n", + "313/313 [==============================] - 13s 42ms/sample - loss: 2.2103 - accuracy: 0.1917\n", "Epoch 2/10\n", - "347/347 [==============================] - 0s 66us/sample - loss: 1.7463 - accuracy: 0.5476\n", + "313/313 [==============================] - 0s 71us/sample - loss: 1.6949 - accuracy: 0.5335\n", "Epoch 3/10\n", - "347/347 [==============================] - 0s 61us/sample - loss: 1.2929 - accuracy: 0.6398\n", + "313/313 [==============================] - 0s 68us/sample - loss: 1.3132 - accuracy: 0.5719\n", "Epoch 4/10\n", - "347/347 [==============================] - 0s 60us/sample - loss: 0.9669 - accuracy: 0.7003\n", + "313/313 [==============================] - 0s 69us/sample - loss: 0.9114 - accuracy: 0.7412\n", "Epoch 5/10\n", - "347/347 [==============================] - 0s 59us/sample - loss: 0.7619 - accuracy: 0.7464\n", + "313/313 [==============================] - 0s 74us/sample - loss: 0.7148 - accuracy: 0.7732\n", "Epoch 6/10\n", - "347/347 [==============================] - 0s 58us/sample - loss: 0.6164 - accuracy: 0.8069\n", + "313/313 [==============================] - 0s 68us/sample - loss: 0.5925 - accuracy: 0.8243\n", "Epoch 7/10\n", - "347/347 [==============================] - 0s 78us/sample - loss: 0.5494 - accuracy: 0.8156\n", + "313/313 [==============================] - 0s 66us/sample - loss: 0.5067 - accuracy: 0.8403\n", "Epoch 8/10\n", - "347/347 [==============================] - 0s 61us/sample - loss: 0.4338 - accuracy: 0.8588\n", + "313/313 [==============================] - 0s 69us/sample - loss: 0.4284 - accuracy: 0.8562\n", "Epoch 9/10\n", - "347/347 [==============================] - 0s 60us/sample - loss: 0.3618 - accuracy: 0.8905\n", + "313/313 [==============================] - 0s 68us/sample - loss: 0.3556 - accuracy: 0.8882\n", "Epoch 10/10\n", - "347/347 [==============================] - 0s 61us/sample - loss: 0.3834 - accuracy: 0.8963\n", - "Train on 344 samples\n", + "313/313 [==============================] - 0s 70us/sample - loss: 0.3052 - accuracy: 0.8914\n", + "Train on 343 samples\n", "Epoch 1/10\n", - "344/344 [==============================] - 1s 2ms/sample - loss: 2.2126 - accuracy: 0.2297\n", + "343/343 [==============================] - 1s 3ms/sample - loss: 2.2128 - accuracy: 0.1924\n", "Epoch 2/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 1.7994 - accuracy: 0.4331\n", + "343/343 [==============================] - 0s 65us/sample - loss: 1.7686 - accuracy: 0.4577\n", "Epoch 3/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 1.3368 - accuracy: 0.5901\n", + "343/343 [==============================] - 0s 66us/sample - loss: 1.3149 - accuracy: 0.6093\n", "Epoch 4/10\n", - "344/344 [==============================] - 0s 67us/sample - loss: 1.0669 - accuracy: 0.6366\n", + "343/343 [==============================] - 0s 67us/sample - loss: 0.9948 - accuracy: 0.6822\n", "Epoch 5/10\n", - "344/344 [==============================] - 0s 63us/sample - loss: 0.8400 - accuracy: 0.7122\n", + "343/343 [==============================] - 0s 63us/sample - loss: 0.8470 - accuracy: 0.7230\n", "Epoch 6/10\n", - "344/344 [==============================] - 0s 62us/sample - loss: 0.6880 - accuracy: 0.8023\n", + "343/343 [==============================] - 0s 64us/sample - loss: 0.7368 - accuracy: 0.7609\n", "Epoch 7/10\n", - "344/344 [==============================] - 0s 63us/sample - loss: 0.6103 - accuracy: 0.7965\n", + "343/343 [==============================] - 0s 65us/sample - loss: 0.6659 - accuracy: 0.7930\n", "Epoch 8/10\n", - "344/344 [==============================] - 0s 60us/sample - loss: 0.5065 - accuracy: 0.8343\n", + "343/343 [==============================] - 0s 65us/sample - loss: 0.5063 - accuracy: 0.8280\n", "Epoch 9/10\n", - "344/344 [==============================] - 0s 64us/sample - loss: 0.4008 - accuracy: 0.8837\n", + "343/343 [==============================] - 0s 65us/sample - loss: 0.4646 - accuracy: 0.8484\n", "Epoch 10/10\n", - "344/344 [==============================] - 0s 61us/sample - loss: 0.3994 - accuracy: 0.8750\n" + "343/343 [==============================] - 0s 71us/sample - loss: 0.4105 - accuracy: 0.8688\n" ] } ], @@ -1942,7 +1970,7 @@ "output_type": "stream", "text": [ "\n", - "Clean test set accuracy (model): 98.21%\n" + "Clean test set accuracy (model): 98.01%\n" ] }, { @@ -1997,7 +2025,7 @@ "output_type": "stream", "text": [ "\n", - "Clean test set accuracy (DPA model_10): 93.99%\n" + "Clean test set accuracy (DPA model_10): 93.94%\n" ] }, { @@ -2052,7 +2080,7 @@ "output_type": "stream", "text": [ "\n", - "Clean test set accuracy (DPA model_20): 91.27%\n" + "Clean test set accuracy (DPA model_20): 91.32%\n" ] }, { @@ -2098,38 +2126,10 @@ }, { "cell_type": "code", - "execution_count": 13, + "execution_count": null, "id": "74d91009", "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "Clean test set accuracy (DPA model_30): 89.95%\n" - ] - }, - { - "data": { - "image/png": "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\n", - "text/plain": [ - "
" - ] - }, - "metadata": { - "needs_background": "light" - }, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Prediction: 0\n" - ] - } - ], + "outputs": [], "source": [ "clean_preds = np.argmax(dpa_model_30.predict(x_test), axis=1)\n", "clean_correct = np.sum(clean_preds == np.argmax(y_test, axis=1))\n", @@ -2162,38 +2162,10 @@ }, { "cell_type": "code", - "execution_count": 14, + "execution_count": null, "id": "1cd59ae2", "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "Poison test set accuracy (model): 0.04%\n" - ] - }, - { - "data": { - "image/png": "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\n", - "text/plain": [ - "
" - ] - }, - "metadata": { - "needs_background": "light" - }, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Prediction: 9\n" - ] - } - ], + "outputs": [], "source": [ "not_target = np.logical_not(np.all(y_test == targets, axis=1))\n", "px_test, py_test = backdoor.poison(x_test[not_target], y_test[not_target])\n", @@ -2214,38 +2186,10 @@ }, { "cell_type": "code", - "execution_count": 15, + "execution_count": null, "id": "500e4154", "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "Poison test set accuracy (DPA model_10): 70.06%\n" - ] - }, - { - "data": { - "image/png": "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\n", - "text/plain": [ - "
" - ] - }, - "metadata": { - "needs_background": "light" - }, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Prediction: 7\n" - ] - } - ], + "outputs": [], "source": [ "poison_preds = np.argmax(dpa_model_10.predict(px_test), axis=1)\n", "clean_correct = np.sum(poison_preds == np.argmax(y_test[not_target], axis=1))\n", @@ -2263,38 +2207,10 @@ }, { "cell_type": "code", - "execution_count": 16, + "execution_count": null, "id": "54e42d16", "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "Poison test set accuracy (DPA model_20): 72.86%\n" - ] - }, - { - "data": { - "image/png": "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\n", - "text/plain": [ - "
" - ] - }, - "metadata": { - "needs_background": "light" - }, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Prediction: 7\n" - ] - } - ], + "outputs": [], "source": [ "poison_preds = np.argmax(dpa_model_20.predict(px_test), axis=1)\n", "clean_correct = np.sum(poison_preds == np.argmax(y_test[not_target], axis=1))\n", @@ -2312,38 +2228,10 @@ }, { "cell_type": "code", - "execution_count": 17, + "execution_count": null, "id": "89dc71c1", "metadata": {}, - "outputs": [ - { - "name": "stdout", - "output_type": "stream", - "text": [ - "\n", - "Poison test set accuracy (DPA model_30): 73.68%\n" - ] - }, - { - "data": { - "image/png": "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\n", - "text/plain": [ - "
" - ] - }, - "metadata": { - "needs_background": "light" - }, - "output_type": "display_data" - }, - { - "name": "stdout", - "output_type": "stream", - "text": [ - "Prediction: 7\n" - ] - } - ], + "outputs": [], "source": [ "poison_preds = np.argmax(dpa_model_30.predict(px_test), axis=1)\n", "clean_correct = np.sum(poison_preds == np.argmax(y_test[not_target], axis=1))\n",