Skip to content
This repository has been archived by the owner on Sep 29, 2020. It is now read-only.

Reduction of the Learning rate isn't working yet #10

Closed
beckstev opened this issue Jun 16, 2019 · 3 comments
Closed

Reduction of the Learning rate isn't working yet #10

beckstev opened this issue Jun 16, 2019 · 3 comments
Labels
invalid This doesn't seem right
Projects
Road to success

Comments

@beckstev
Copy link
Owner

We should implement in future version a dynamic leraning rate. See the following post:

keras-team/keras#2823
keras-team/keras#898
@beckstev beckstev added the enhancement New feature or request label Jun 16, 2019
@beckstev
Copy link
Owner Author

I implemented EarlyStopping and AdaptiveLeariningRate.
We should also check out Keras Callback functions.

@beckstev beckstev added the bug Something isn't working label Jun 16, 2019
@beckstev beckstev changed the title Adapt the learning rate during the traning Reduction of the Learning rate isn't working yet Jun 16, 2019
@beckstev
Copy link
Owner Author

So I trained DogNNv2 for 2.5 hours and you can see the loss is decreasing 👎
I noticed that I forgot to activate EarlyStopping and LRreducer 👎

720/720 [==============================] - 533s 740ms/step - loss: 4.7865 - acc: 0.0102 - val_loss: 4.7853 - val_acc: 0.0124
Epoch 2/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7828 - acc: 0.0109 - val_loss: 4.7845 - val_acc: 0.0124
Epoch 3/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7806 - acc: 0.0115 - val_loss: 4.7846 - val_acc: 0.0122
Epoch 4/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7798 - acc: 0.0119 - val_loss: 4.7850 - val_acc: 0.0124
Epoch 5/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7794 - acc: 0.0114 - val_loss: 4.7854 - val_acc: 0.0122
Epoch 6/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7791 - acc: 0.0109 - val_loss: 4.7858 - val_acc: 0.0122
Epoch 7/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7790 - acc: 0.0115 - val_loss: 4.7854 - val_acc: 0.0122
Epoch 8/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7790 - acc: 0.0117 - val_loss: 4.7857 - val_acc: 0.0122
Epoch 9/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7789 - acc: 0.0120 - val_loss: 4.7861 - val_acc: 0.0122
Epoch 10/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7789 - acc: 0.0120 - val_loss: 4.7861 - val_acc: 0.0122
Epoch 11/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7789 - acc: 0.0115 - val_loss: 4.7860 - val_acc: 0.0122
Epoch 12/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7788 - acc: 0.0120 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 13/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7789 - acc: 0.0111 - val_loss: 4.7861 - val_acc: 0.0122
Epoch 14/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7788 - acc: 0.0120 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 15/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7788 - acc: 0.0117 - val_loss: 4.7862 - val_acc: 0.0122
Epoch 16/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7787 - acc: 0.0120 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 17/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7787 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 18/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7787 - acc: 0.0116 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 19/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7787 - acc: 0.0112 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 20/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7787 - acc: 0.0115 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 21/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7787 - acc: 0.0120 - val_loss: 4.7862 - val_acc: 0.0122
Epoch 22/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7787 - acc: 0.0120 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 23/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0110 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 24/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0099 - val_loss: 4.7862 - val_acc: 0.0122
Epoch 25/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0113 - val_loss: 4.7862 - val_acc: 0.0122
Epoch 26/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0117 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 27/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0116 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 28/10000
720/720 [==============================] - 94s 130ms/step - loss: 4.7786 - acc: 0.0120 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 29/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0120 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 30/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0115 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 31/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0110 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 32/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0120 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 33/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7786 - acc: 0.0120 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 34/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0105 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 35/10000
720/720 [==============================] - 94s 130ms/step - loss: 4.7785 - acc: 0.0116 - val_loss: 4.7863 - val_acc: 0.0122
Epoch 36/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0117 - val_loss: 4.7864 - val_acc: 0.0122
Epoch 37/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7786 - acc: 0.0117 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 38/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0109 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 39/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0115 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 40/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0117 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 41/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7785 - acc: 0.0116 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 42/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 43/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 44/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 45/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 46/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0109 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 47/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0111 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 48/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 49/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0104 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 50/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 51/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0117 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 52/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 53/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0114 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 54/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0106 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 55/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0111 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 56/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0109 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 57/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 58/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0111 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 59/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 60/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0111 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 61/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 62/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0119 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 63/10000
720/720 [==============================] - 95s 131ms/step - loss: 4.7785 - acc: 0.0108 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 64/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0115 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 65/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 66/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 67/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 68/10000
720/720 [==============================] - 94s 130ms/step - loss: 4.7785 - acc: 0.0109 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 69/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 70/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0115 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 71/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0115 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 72/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0116 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 73/10000
720/720 [==============================] - 94s 130ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 74/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0113 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 75/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0114 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 76/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 77/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0116 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 78/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 79/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 80/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 81/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0113 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 82/10000
720/720 [==============================] - 93s 130ms/step - loss: 4.7785 - acc: 0.0116 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 83/10000
720/720 [==============================] - 94s 130ms/step - loss: 4.7785 - acc: 0.0112 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 84/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0112 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 85/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0114 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 86/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7865 - val_acc: 0.0122
Epoch 87/10000
720/720 [==============================] - 95s 132ms/step - loss: 4.7785 - acc: 0.0115 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 88/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 89/10000
720/720 [==============================] - 94s 130ms/step - loss: 4.7785 - acc: 0.0114 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 90/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 91/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0108 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 92/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0120 - val_loss: 4.7866 - val_acc: 0.0122
Epoch 93/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0119 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 94/10000
720/720 [==============================] - 94s 131ms/step - loss: 4.7785 - acc: 0.0113 - val_loss: 4.7867 - val_acc: 0.0122
Epoch 95/10000

@beckstev beckstev added invalid This doesn't seem right and removed bug Something isn't working enhancement New feature or request labels Jun 16, 2019
@beckstev
Copy link
Owner Author

Implemented a functional solution in #13

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees
No one assigned
Labels
invalid This doesn't seem right
Projects
Road to success
  
Awaiting triage
Milestone
No milestone
Development

No branches or pull requests
1 participant
@beckstev