4_convolutions.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "4embtkV0pNxM"
   },
   "source": [
    "Deep Learning\n",
    "=============\n",
    "\n",
    "Assignment 4\n",
    "------------\n",
    "\n",
    "Previously in `2_fullyconnected.ipynb` and `3_regularization.ipynb`, we trained fully connected networks to classify [notMNIST](http://yaroslavvb.blogspot.com/2011/09/notmnist-dataset.html) characters.\n",
    "\n",
    "The goal of this assignment is make the neural network convolutional."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "collapsed": true,
    "id": "tm2CQN_Cpwj0"
   },
   "outputs": [],
   "source": [
    "# These are all the modules we'll be using later. Make sure you can import them\n",
    "# before proceeding further.\n",
    "from __future__ import print_function\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from six.moves import cPickle as pickle\n",
    "from six.moves import range"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "collapsed": false,
    "executionInfo": {
     "elapsed": 11948,
     "status": "ok",
     "timestamp": 1446658914837,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "y3-cj1bpmuxc",
    "outputId": "016b1a51-0290-4b08-efdb-8c95ffc3cd01"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set (504090, 28, 28) (504090,)\n",
      "Validation set (25000, 28, 28) (25000,)\n",
      "Test set (18720, 28, 28) (18720,)\n"
     ]
    }
   ],
   "source": [
    "pickle_file = 'notMNIST.pickle'\n",
    "\n",
    "with open(pickle_file, 'rb') as f:\n",
    "  save = pickle.load(f)\n",
    "  train_dataset = save['train_dataset']\n",
    "  train_labels = save['train_labels']\n",
    "  valid_dataset = save['valid_dataset']\n",
    "  valid_labels = save['valid_labels']\n",
    "  test_dataset = save['test_dataset']\n",
    "  test_labels = save['test_labels']\n",
    "  del save  # hint to help gc free up memory\n",
    "  print('Training set', train_dataset.shape, train_labels.shape)\n",
    "  print('Validation set', valid_dataset.shape, valid_labels.shape)\n",
    "  print('Test set', test_dataset.shape, test_labels.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "L7aHrm6nGDMB"
   },
   "source": [
    "Reformat into a TensorFlow-friendly shape:\n",
    "- convolutions need the image data formatted as a cube (width by height by #channels)\n",
    "- labels as float 1-hot encodings."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "collapsed": false,
    "executionInfo": {
     "elapsed": 11952,
     "status": "ok",
     "timestamp": 1446658914857,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "IRSyYiIIGIzS",
    "outputId": "650a208c-8359-4852-f4f5-8bf10e80ef6c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set (504090, 28, 28, 1) (504090, 10)\n",
      "Validation set (25000, 28, 28, 1) (25000, 10)\n",
      "Test set (18720, 28, 28, 1) (18720, 10)\n"
     ]
    }
   ],
   "source": [
    "image_size = 28\n",
    "num_labels = 10\n",
    "num_channels = 1 # grayscale\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "def reformat(dataset, labels):\n",
    "  dataset = dataset.reshape(\n",
    "    (-1, image_size, image_size, num_channels)).astype(np.float32)\n",
    "  labels = (np.arange(num_labels) == labels[:,None]).astype(np.float32)\n",
    "  return dataset, labels\n",
    "train_dataset, train_labels = reformat(train_dataset, train_labels)\n",
    "valid_dataset, valid_labels = reformat(valid_dataset, valid_labels)\n",
    "test_dataset, test_labels = reformat(test_dataset, test_labels)\n",
    "print('Training set', train_dataset.shape, train_labels.shape)\n",
    "print('Validation set', valid_dataset.shape, valid_labels.shape)\n",
    "print('Test set', test_dataset.shape, test_labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "collapsed": true,
    "id": "AgQDIREv02p1"
   },
   "outputs": [],
   "source": [
    "def accuracy(predictions, labels):\n",
    "  return (100.0 * np.sum(np.argmax(predictions, 1) == np.argmax(labels, 1))\n",
    "          / predictions.shape[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "5rhgjmROXu2O"
   },
   "source": [
    "Let's build a small network with two convolutional layers, followed by one fully connected layer. Convolutional networks are more expensive computationally, so we'll limit its depth and number of fully connected nodes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "collapsed": true,
    "id": "IZYv70SvvOan"
   },
   "outputs": [],
   "source": [
    "batch_size = 16\n",
    "patch_size = 5\n",
    "depth = 16\n",
    "num_hidden = 64\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default():\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "  \n",
    "  # Variables.\n",
    "  layer1_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
    "  layer1_biases = tf.Variable(tf.zeros([depth]))\n",
    "  layer2_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, depth, depth], stddev=0.1))\n",
    "  layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
    "  layer3_weights = tf.Variable(tf.truncated_normal(\n",
    "      [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
    "  layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
    "  layer4_weights = tf.Variable(tf.truncated_normal(\n",
    "      [num_hidden, num_labels], stddev=0.1))\n",
    "  layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "  \n",
    "  # Model.\n",
    "  def model(data):\n",
    "    conv = tf.nn.conv2d(data, layer1_weights, [1, 2, 2, 1], padding='SAME')\n",
    "    hidden = tf.nn.relu(conv + layer1_biases)\n",
    "    conv = tf.nn.conv2d(hidden, layer2_weights, [1, 2, 2, 1], padding='SAME')\n",
    "    hidden = tf.nn.relu(conv + layer2_biases)\n",
    "    shape = hidden.get_shape().as_list()\n",
    "    reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "    return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset)\n",
    "  loss = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(logits, tf_train_labels))\n",
    "    \n",
    "  # Optimizer.\n",
    "  optimizer = tf.train.GradientDescentOptimizer(0.05).minimize(loss)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "cellView": "both",
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "output_extras": [
      {
       "item_id": 37
      }
     ]
    },
    "colab_type": "code",
    "collapsed": false,
    "executionInfo": {
     "elapsed": 63292,
     "status": "ok",
     "timestamp": 1446658966251,
     "user": {
      "color": "",
      "displayName": "",
      "isAnonymous": false,
      "isMe": true,
      "permissionId": "",
      "photoUrl": "",
      "sessionId": "0",
      "userId": ""
     },
     "user_tz": 480
    },
    "id": "noKFb2UovVFR",
    "outputId": "28941338-2ef9-4088-8bd1-44295661e628"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized\n",
      "Minibatch loss at step 0: 4.517112\n",
      "Minibatch accuracy: 12.5%\n",
      "Validation accuracy: 10.8%\n",
      "Minibatch loss at step 50: 1.901366\n",
      "Minibatch accuracy: 50.0%\n",
      "Validation accuracy: 51.5%\n",
      "Minibatch loss at step 100: 0.960779\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 70.1%\n",
      "Minibatch loss at step 150: 1.263199\n",
      "Minibatch accuracy: 62.5%\n",
      "Validation accuracy: 75.4%\n",
      "Minibatch loss at step 200: 0.801022\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 76.9%\n",
      "Minibatch loss at step 250: 0.716393\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 77.0%\n",
      "Minibatch loss at step 300: 0.562906\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 79.0%\n",
      "Minibatch loss at step 350: 0.608726\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 79.5%\n",
      "Minibatch loss at step 400: 0.664071\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 80.3%\n",
      "Minibatch loss at step 450: 0.084007\n",
      "Minibatch accuracy: 100.0%\n",
      "Validation accuracy: 81.1%\n",
      "Minibatch loss at step 500: 0.808882\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 79.9%\n",
      "Minibatch loss at step 550: 0.231459\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 81.6%\n",
      "Minibatch loss at step 600: 0.613433\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 81.4%\n",
      "Minibatch loss at step 650: 0.581551\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 82.3%\n",
      "Minibatch loss at step 700: 0.293391\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 82.4%\n",
      "Minibatch loss at step 750: 0.247936\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 82.2%\n",
      "Minibatch loss at step 800: 1.182951\n",
      "Minibatch accuracy: 56.2%\n",
      "Validation accuracy: 82.7%\n",
      "Minibatch loss at step 850: 0.577485\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 82.5%\n",
      "Minibatch loss at step 900: 0.614829\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 82.9%\n",
      "Minibatch loss at step 950: 1.228545\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 82.8%\n",
      "Minibatch loss at step 1000: 0.424906\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 82.4%\n",
      "Test accuracy: 89.3%\n"
     ]
    }
   ],
   "source": [
    "num_steps = 1001\n",
    "\n",
    "with tf.Session(graph=graph) as session:\n",
    "  tf.initialize_all_variables().run()\n",
    "  print('Initialized')\n",
    "  for step in range(num_steps):\n",
    "    offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "    batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "    batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "    feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "    _, l, predictions = session.run(\n",
    "      [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "    if (step % 50 == 0):\n",
    "      print('Minibatch loss at step %d: %f' % (step, l))\n",
    "      print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "      print('Validation accuracy: %.1f%%' % accuracy(\n",
    "        valid_prediction.eval(), valid_labels))\n",
    "  print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "KedKkn4EutIK"
   },
   "source": [
    "---\n",
    "Problem 1\n",
    "---------\n",
    "\n",
    "The convolutional model above uses convolutions with stride 2 to reduce the dimensionality. Replace the strides by a max pooling operation (`nn.max_pool()`) of stride 2 and kernel size 2.\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized\n",
      "Minibatch loss at step 0: 3.203771\n",
      "Minibatch accuracy: 12.5%\n",
      "Validation accuracy: 9.9%\n",
      "Minibatch loss at step 50: 1.894837\n",
      "Minibatch accuracy: 43.8%\n",
      "Validation accuracy: 47.9%\n",
      "Minibatch loss at step 100: 1.133459\n",
      "Minibatch accuracy: 50.0%\n",
      "Validation accuracy: 63.4%\n",
      "Minibatch loss at step 150: 1.372210\n",
      "Minibatch accuracy: 62.5%\n",
      "Validation accuracy: 70.2%\n",
      "Minibatch loss at step 200: 0.962181\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 73.4%\n",
      "Minibatch loss at step 250: 0.808819\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 76.4%\n",
      "Minibatch loss at step 300: 0.720418\n",
      "Minibatch accuracy: 75.0%\n",
      "Validation accuracy: 79.7%\n",
      "Minibatch loss at step 350: 0.646657\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 79.7%\n",
      "Minibatch loss at step 400: 0.584711\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 80.8%\n",
      "Minibatch loss at step 450: 0.102403\n",
      "Minibatch accuracy: 100.0%\n",
      "Validation accuracy: 81.6%\n",
      "Minibatch loss at step 500: 0.752791\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 80.2%\n",
      "Minibatch loss at step 550: 0.142285\n",
      "Minibatch accuracy: 100.0%\n",
      "Validation accuracy: 82.4%\n",
      "Minibatch loss at step 600: 0.723146\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 80.1%\n",
      "Minibatch loss at step 650: 0.506009\n",
      "Minibatch accuracy: 87.5%\n",
      "Validation accuracy: 82.9%\n",
      "Minibatch loss at step 700: 0.262775\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 82.7%\n",
      "Minibatch loss at step 750: 0.179683\n",
      "Minibatch accuracy: 93.8%\n",
      "Validation accuracy: 82.5%\n",
      "Minibatch loss at step 800: 1.132485\n",
      "Minibatch accuracy: 56.2%\n",
      "Validation accuracy: 83.7%\n",
      "Minibatch loss at step 850: 0.574637\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 83.2%\n",
      "Minibatch loss at step 900: 0.715934\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 83.9%\n",
      "Minibatch loss at step 950: 1.201541\n",
      "Minibatch accuracy: 68.8%\n",
      "Validation accuracy: 83.1%\n",
      "Minibatch loss at step 1000: 0.614954\n",
      "Minibatch accuracy: 81.2%\n",
      "Validation accuracy: 83.4%\n",
      "Test accuracy: 89.9%\n"
     ]
    }
   ],
   "source": [
    "batch_size = 16\n",
    "patch_size = 5\n",
    "depth = 16\n",
    "num_hidden = 64\n",
    "num_steps = 1001\n",
    "\n",
    "graph = tf.Graph()\n",
    "\n",
    "with graph.as_default():\n",
    "\n",
    "  # Input data.\n",
    "  tf_train_dataset = tf.placeholder(\n",
    "    tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "  tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "  tf_valid_dataset = tf.constant(valid_dataset)\n",
    "  tf_test_dataset = tf.constant(test_dataset)\n",
    "  \n",
    "  # Variables.\n",
    "  layer1_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
    "  layer1_biases = tf.Variable(tf.zeros([depth]))\n",
    "  layer2_weights = tf.Variable(tf.truncated_normal(\n",
    "      [patch_size, patch_size, depth, depth], stddev=0.1))\n",
    "  layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
    "  layer3_weights = tf.Variable(tf.truncated_normal(\n",
    "      [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
    "  layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
    "  layer4_weights = tf.Variable(tf.truncated_normal(\n",
    "      [num_hidden, num_labels], stddev=0.1))\n",
    "  layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "  \n",
    "  # Model.\n",
    "  def model(data):\n",
    "    conv = tf.nn.conv2d(data, layer1_weights, [1, 1, 1, 1], padding='SAME')\n",
    "    relu = tf.nn.relu(conv + layer1_biases)\n",
    "    hidden = tf.nn.max_pool(relu, [1,2,2,1], [1,2,2,1], padding='SAME')\n",
    "    conv = tf.nn.conv2d(hidden, layer2_weights, [1, 1, 1, 1], padding='SAME')\n",
    "    relu = tf.nn.relu(conv + layer2_biases)\n",
    "    hidden = tf.nn.max_pool(relu, [1,2,2,1], [1,2,2,1], padding='SAME')\n",
    "    shape = hidden.get_shape().as_list()\n",
    "    reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "    hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "    return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
    "  \n",
    "  # Training computation.\n",
    "  logits = model(tf_train_dataset)\n",
    "  loss = tf.reduce_mean(\n",
    "    tf.nn.softmax_cross_entropy_with_logits(logits, tf_train_labels))\n",
    "    \n",
    "  # Optimizer.\n",
    "  optimizer = tf.train.GradientDescentOptimizer(0.05).minimize(loss)\n",
    "  \n",
    "  # Predictions for the training, validation, and test data.\n",
    "  train_prediction = tf.nn.softmax(logits)\n",
    "  valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "  test_prediction = tf.nn.softmax(model(tf_test_dataset))\n",
    "\n",
    "\n",
    "with tf.Session(graph=graph) as session:\n",
    "  tf.initialize_all_variables().run()\n",
    "  print('Initialized')\n",
    "  for step in range(num_steps):\n",
    "    offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "    batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "    batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "    feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "    _, l, predictions = session.run(\n",
    "      [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "    if (step % 50 == 0):\n",
    "      print('Minibatch loss at step %d: %f' % (step, l))\n",
    "      print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "      print('Validation accuracy: %.1f%%' % accuracy(\n",
    "        valid_prediction.eval(), valid_labels))\n",
    "  print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "klf21gpbAgb-"
   },
   "source": [
    "---\n",
    "Problem 2\n",
    "---------\n",
    "\n",
    "Try to get the best performance you can using a convolutional net. Look for example at the classic [LeNet5](http://yann.lecun.com/exdb/lenet/) architecture, adding Dropout, and/or adding learning rate decay.\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized kr: 0.95, ilr: 0.8, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.8, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.8, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.8, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.8, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.4, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.4, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.4, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.4, drv: 0.9\n",
      "Test accuracy: 29.3%\n",
      "Initialized kr: 0.95, ilr: 0.4, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.6\n",
      "Test accuracy: 86.9%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.7\n",
      "Test accuracy: 85.7%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.8\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 91.2%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.6\n",
      "Test accuracy: 86.7%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.7\n",
      "Test accuracy: 86.9%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.8\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 91.2%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.6\n",
      "Test accuracy: 82.1%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.7\n",
      "Test accuracy: 86.6%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.8\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.95, ilr: 0.025, drv: 0.6\n",
      "Test accuracy: 81.0%\n",
      "Initialized kr: 0.95, ilr: 0.025, drv: 0.7\n",
      "Test accuracy: 84.0%\n",
      "Initialized kr: 0.95, ilr: 0.025, drv: 0.8\n",
      "Test accuracy: 87.0%\n",
      "Initialized kr: 0.95, ilr: 0.025, drv: 0.9\n",
      "Test accuracy: 86.6%\n",
      "Initialized kr: 0.95, ilr: 0.025, drv: 0.95\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.95, ilr: 0.0125, drv: 0.6\n",
      "Test accuracy: 67.0%\n",
      "Initialized kr: 0.95, ilr: 0.0125, drv: 0.7\n",
      "Test accuracy: 78.0%\n",
      "Initialized kr: 0.95, ilr: 0.0125, drv: 0.8\n",
      "Test accuracy: 84.4%\n",
      "Initialized kr: 0.95, ilr: 0.0125, drv: 0.9\n",
      "Test accuracy: 85.2%\n",
      "Initialized kr: 0.95, ilr: 0.0125, drv: 0.95\n",
      "Test accuracy: 86.5%\n",
      "Initialized kr: 0.95, ilr: 0.00625, drv: 0.6\n",
      "Test accuracy: 49.6%\n",
      "Initialized kr: 0.95, ilr: 0.00625, drv: 0.7\n",
      "Test accuracy: 58.4%\n",
      "Initialized kr: 0.95, ilr: 0.00625, drv: 0.8\n",
      "Test accuracy: 79.0%\n",
      "Initialized kr: 0.95, ilr: 0.00625, drv: 0.9\n",
      "Test accuracy: 83.6%\n",
      "Initialized kr: 0.95, ilr: 0.00625, drv: 0.95\n",
      "Test accuracy: 84.6%\n",
      "Initialized kr: 0.95, ilr: 0.003125, drv: 0.6\n",
      "Test accuracy: 23.6%\n",
      "Initialized kr: 0.95, ilr: 0.003125, drv: 0.7\n",
      "Test accuracy: 42.7%\n",
      "Initialized kr: 0.95, ilr: 0.003125, drv: 0.8\n",
      "Test accuracy: 59.2%\n",
      "Initialized kr: 0.95, ilr: 0.003125, drv: 0.9\n",
      "Test accuracy: 69.9%\n",
      "Initialized kr: 0.95, ilr: 0.003125, drv: 0.95\n",
      "Test accuracy: 77.1%\n",
      "Initialized kr: 0.875, ilr: 0.8, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.8, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.8, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.8, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.8, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.4, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.4, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.4, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.4, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.4, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.2, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.2, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.2, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.875, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.875, ilr: 0.1, drv: 0.6\n",
      "Test accuracy: 86.7%\n",
      "Initialized kr: 0.875, ilr: 0.1, drv: 0.7\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.875, ilr: 0.1, drv: 0.8\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.875, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.875, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.875, ilr: 0.05, drv: 0.6\n",
      "Test accuracy: 85.9%\n",
      "Initialized kr: 0.875, ilr: 0.05, drv: 0.7\n",
      "Test accuracy: 85.9%\n",
      "Initialized kr: 0.875, ilr: 0.05, drv: 0.8\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.875, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.875, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 90.1%\n",
      "Initialized kr: 0.875, ilr: 0.025, drv: 0.6\n",
      "Test accuracy: 80.8%\n",
      "Initialized kr: 0.875, ilr: 0.025, drv: 0.7\n",
      "Test accuracy: 85.2%\n",
      "Initialized kr: 0.875, ilr: 0.025, drv: 0.8\n",
      "Test accuracy: 86.1%\n",
      "Initialized kr: 0.875, ilr: 0.025, drv: 0.9\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.875, ilr: 0.025, drv: 0.95\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.875, ilr: 0.0125, drv: 0.6\n",
      "Test accuracy: 72.0%\n",
      "Initialized kr: 0.875, ilr: 0.0125, drv: 0.7\n",
      "Test accuracy: 81.5%\n",
      "Initialized kr: 0.875, ilr: 0.0125, drv: 0.8\n",
      "Test accuracy: 83.9%\n",
      "Initialized kr: 0.875, ilr: 0.0125, drv: 0.9\n",
      "Test accuracy: 85.6%\n",
      "Initialized kr: 0.875, ilr: 0.0125, drv: 0.95\n",
      "Test accuracy: 86.5%\n",
      "Initialized kr: 0.875, ilr: 0.00625, drv: 0.6\n",
      "Test accuracy: 49.0%\n",
      "Initialized kr: 0.875, ilr: 0.00625, drv: 0.7\n",
      "Test accuracy: 68.3%\n",
      "Initialized kr: 0.875, ilr: 0.00625, drv: 0.8\n",
      "Test accuracy: 79.7%\n",
      "Initialized kr: 0.875, ilr: 0.00625, drv: 0.9\n",
      "Test accuracy: 83.1%\n",
      "Initialized kr: 0.875, ilr: 0.00625, drv: 0.95\n",
      "Test accuracy: 84.4%\n",
      "Initialized kr: 0.875, ilr: 0.003125, drv: 0.6\n",
      "Test accuracy: 33.6%\n",
      "Initialized kr: 0.875, ilr: 0.003125, drv: 0.7\n",
      "Test accuracy: 35.7%\n",
      "Initialized kr: 0.875, ilr: 0.003125, drv: 0.8\n",
      "Test accuracy: 61.2%\n",
      "Initialized kr: 0.875, ilr: 0.003125, drv: 0.9\n",
      "Test accuracy: 79.7%\n",
      "Initialized kr: 0.875, ilr: 0.003125, drv: 0.95\n",
      "Test accuracy: 82.6%\n",
      "Initialized kr: 0.75, ilr: 0.8, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.8, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.8, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.8, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.8, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.4, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.4, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.4, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.4, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.4, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.8\n",
      "Test accuracy: 85.6%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.6\n",
      "Test accuracy: 86.9%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.7\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.8\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.9%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 89.9%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.6\n",
      "Test accuracy: 85.1%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.7\n",
      "Test accuracy: 86.2%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.8\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.75, ilr: 0.025, drv: 0.6\n",
      "Test accuracy: 81.5%\n",
      "Initialized kr: 0.75, ilr: 0.025, drv: 0.7\n",
      "Test accuracy: 83.9%\n",
      "Initialized kr: 0.75, ilr: 0.025, drv: 0.8\n",
      "Test accuracy: 86.3%\n",
      "Initialized kr: 0.75, ilr: 0.025, drv: 0.9\n",
      "Test accuracy: 87.2%\n",
      "Initialized kr: 0.75, ilr: 0.025, drv: 0.95\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.75, ilr: 0.0125, drv: 0.6\n",
      "Test accuracy: 78.0%\n",
      "Initialized kr: 0.75, ilr: 0.0125, drv: 0.7\n",
      "Test accuracy: 78.8%\n",
      "Initialized kr: 0.75, ilr: 0.0125, drv: 0.8\n",
      "Test accuracy: 83.1%\n",
      "Initialized kr: 0.75, ilr: 0.0125, drv: 0.9\n",
      "Test accuracy: 84.8%\n",
      "Initialized kr: 0.75, ilr: 0.0125, drv: 0.95\n",
      "Test accuracy: 86.4%\n",
      "Initialized kr: 0.75, ilr: 0.00625, drv: 0.6\n",
      "Test accuracy: 56.9%\n",
      "Initialized kr: 0.75, ilr: 0.00625, drv: 0.7\n",
      "Test accuracy: 68.7%\n",
      "Initialized kr: 0.75, ilr: 0.00625, drv: 0.8\n",
      "Test accuracy: 79.7%\n",
      "Initialized kr: 0.75, ilr: 0.00625, drv: 0.9\n",
      "Test accuracy: 84.0%\n",
      "Initialized kr: 0.75, ilr: 0.00625, drv: 0.95\n",
      "Test accuracy: 84.9%\n",
      "Initialized kr: 0.75, ilr: 0.003125, drv: 0.6\n",
      "Test accuracy: 35.3%\n",
      "Initialized kr: 0.75, ilr: 0.003125, drv: 0.7\n",
      "Test accuracy: 44.2%\n",
      "Initialized kr: 0.75, ilr: 0.003125, drv: 0.8\n",
      "Test accuracy: 70.5%\n",
      "Initialized kr: 0.75, ilr: 0.003125, drv: 0.9\n",
      "Test accuracy: 79.0%\n",
      "Initialized kr: 0.75, ilr: 0.003125, drv: 0.95\n",
      "Test accuracy: 80.8%\n",
      "Initialized kr: 0.5, ilr: 0.8, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.8, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.8, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.8, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.8, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.4, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.4, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.4, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.4, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.4, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.2, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.2, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.2, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.5, ilr: 0.1, drv: 0.6\n",
      "Test accuracy: 84.2%\n",
      "Initialized kr: 0.5, ilr: 0.1, drv: 0.7\n",
      "Test accuracy: 83.8%\n",
      "Initialized kr: 0.5, ilr: 0.1, drv: 0.8\n",
      "Test accuracy: 86.3%\n",
      "Initialized kr: 0.5, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.5, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 20.5%\n",
      "Initialized kr: 0.5, ilr: 0.05, drv: 0.6\n",
      "Test accuracy: 81.9%\n",
      "Initialized kr: 0.5, ilr: 0.05, drv: 0.7\n",
      "Test accuracy: 83.9%\n",
      "Initialized kr: 0.5, ilr: 0.05, drv: 0.8\n",
      "Test accuracy: 86.6%\n",
      "Initialized kr: 0.5, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.5, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.5, ilr: 0.025, drv: 0.6\n",
      "Test accuracy: 78.4%\n",
      "Initialized kr: 0.5, ilr: 0.025, drv: 0.7\n",
      "Test accuracy: 82.0%\n",
      "Initialized kr: 0.5, ilr: 0.025, drv: 0.8\n",
      "Test accuracy: 83.9%\n",
      "Initialized kr: 0.5, ilr: 0.025, drv: 0.9\n",
      "Test accuracy: 86.4%\n",
      "Initialized kr: 0.5, ilr: 0.025, drv: 0.95\n",
      "Test accuracy: 86.8%\n",
      "Initialized kr: 0.5, ilr: 0.0125, drv: 0.6\n",
      "Test accuracy: 70.7%\n",
      "Initialized kr: 0.5, ilr: 0.0125, drv: 0.7\n",
      "Test accuracy: 73.6%\n",
      "Initialized kr: 0.5, ilr: 0.0125, drv: 0.8\n",
      "Test accuracy: 80.3%\n",
      "Initialized kr: 0.5, ilr: 0.0125, drv: 0.9\n",
      "Test accuracy: 83.9%\n",
      "Initialized kr: 0.5, ilr: 0.0125, drv: 0.95\n",
      "Test accuracy: 83.6%\n",
      "Initialized kr: 0.5, ilr: 0.00625, drv: 0.6\n",
      "Test accuracy: 62.4%\n",
      "Initialized kr: 0.5, ilr: 0.00625, drv: 0.7\n",
      "Test accuracy: 65.6%\n",
      "Initialized kr: 0.5, ilr: 0.00625, drv: 0.8\n",
      "Test accuracy: 70.8%\n",
      "Initialized kr: 0.5, ilr: 0.00625, drv: 0.9\n",
      "Test accuracy: 80.1%\n",
      "Initialized kr: 0.5, ilr: 0.00625, drv: 0.95\n",
      "Test accuracy: 80.6%\n",
      "Initialized kr: 0.5, ilr: 0.003125, drv: 0.6\n",
      "Test accuracy: 34.3%\n",
      "Initialized kr: 0.5, ilr: 0.003125, drv: 0.7\n",
      "Test accuracy: 40.1%\n",
      "Initialized kr: 0.5, ilr: 0.003125, drv: 0.8\n",
      "Test accuracy: 59.6%\n",
      "Initialized kr: 0.5, ilr: 0.003125, drv: 0.9\n",
      "Test accuracy: 74.1%\n",
      "Initialized kr: 0.5, ilr: 0.003125, drv: 0.95\n",
      "Test accuracy: 77.1%\n",
      "Initialized kr: 0.25, ilr: 0.8, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.8, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.8, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.8, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.8, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.4, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.4, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.4, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.4, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.4, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.2, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.2, drv: 0.7\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.2, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.1, drv: 0.6\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.1, drv: 0.7\n",
      "Test accuracy: 74.9%\n",
      "Initialized kr: 0.25, ilr: 0.1, drv: 0.8\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.25, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 10.1%\n",
      "Initialized kr: 0.25, ilr: 0.05, drv: 0.6\n",
      "Test accuracy: 71.2%\n",
      "Initialized kr: 0.25, ilr: 0.05, drv: 0.7\n",
      "Test accuracy: 68.3%\n",
      "Initialized kr: 0.25, ilr: 0.05, drv: 0.8\n",
      "Test accuracy: 80.2%\n",
      "Initialized kr: 0.25, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 82.1%\n",
      "Initialized kr: 0.25, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 78.8%\n",
      "Initialized kr: 0.25, ilr: 0.025, drv: 0.6\n",
      "Test accuracy: 54.6%\n",
      "Initialized kr: 0.25, ilr: 0.025, drv: 0.7\n",
      "Test accuracy: 64.5%\n",
      "Initialized kr: 0.25, ilr: 0.025, drv: 0.8\n",
      "Test accuracy: 59.4%\n",
      "Initialized kr: 0.25, ilr: 0.025, drv: 0.9\n",
      "Test accuracy: 79.3%\n",
      "Initialized kr: 0.25, ilr: 0.025, drv: 0.95\n",
      "Test accuracy: 82.9%\n",
      "Initialized kr: 0.25, ilr: 0.0125, drv: 0.6\n",
      "Test accuracy: 45.5%\n",
      "Initialized kr: 0.25, ilr: 0.0125, drv: 0.7\n",
      "Test accuracy: 41.1%\n",
      "Initialized kr: 0.25, ilr: 0.0125, drv: 0.8\n",
      "Test accuracy: 66.5%\n",
      "Initialized kr: 0.25, ilr: 0.0125, drv: 0.9\n",
      "Test accuracy: 76.4%\n",
      "Initialized kr: 0.25, ilr: 0.0125, drv: 0.95\n",
      "Test accuracy: 75.8%\n",
      "Initialized kr: 0.25, ilr: 0.00625, drv: 0.6\n",
      "Test accuracy: 34.7%\n",
      "Initialized kr: 0.25, ilr: 0.00625, drv: 0.7\n",
      "Test accuracy: 40.7%\n",
      "Initialized kr: 0.25, ilr: 0.00625, drv: 0.8\n",
      "Test accuracy: 44.4%\n",
      "Initialized kr: 0.25, ilr: 0.00625, drv: 0.9\n",
      "Test accuracy: 53.4%\n",
      "Initialized kr: 0.25, ilr: 0.00625, drv: 0.95\n",
      "Test accuracy: 69.5%\n",
      "Initialized kr: 0.25, ilr: 0.003125, drv: 0.6\n",
      "Test accuracy: 19.3%\n",
      "Initialized kr: 0.25, ilr: 0.003125, drv: 0.7\n",
      "Test accuracy: 22.5%\n",
      "Initialized kr: 0.25, ilr: 0.003125, drv: 0.8\n",
      "Test accuracy: 35.7%\n",
      "Initialized kr: 0.25, ilr: 0.003125, drv: 0.9\n",
      "Test accuracy: 54.7%\n",
      "Initialized kr: 0.25, ilr: 0.003125, drv: 0.95\n",
      "Test accuracy: 60.1%\n"
     ]
    }
   ],
   "source": [
    "for keep_rate in [0.95, 0.875, 0.75, 0.5, 0.25]:\n",
    "    for initial_learning_rate in [0.8, 0.4, 0.2, 0.1, 0.05, 0.025, 0.0125, 0.00625, 0.003125]:\n",
    "        for decay_rate_value in [0.6, 0.7, 0.8, 0.9, 0.95]:\n",
    "\n",
    "            graph = tf.Graph()\n",
    "\n",
    "            with graph.as_default():\n",
    "\n",
    "              # Input data.\n",
    "              tf_train_dataset = tf.placeholder(\n",
    "                tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "              tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "              tf_valid_dataset = tf.constant(valid_dataset)\n",
    "              tf_test_dataset = tf.constant(test_dataset)\n",
    "              \n",
    "              # Variables.\n",
    "              layer1_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
    "              layer1_biases = tf.Variable(tf.zeros([depth]))\n",
    "              layer2_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [patch_size, patch_size, depth, depth], stddev=0.1))\n",
    "              layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
    "              layer3_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
    "              layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
    "              layer4_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [num_hidden, num_labels], stddev=0.1))\n",
    "              layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "              \n",
    "              # Model.\n",
    "              # try 1, 2 more layers\n",
    "              def model(data):\n",
    "                conv = tf.nn.conv2d(data, layer1_weights, [1, 1, 1, 1], padding='SAME')\n",
    "                relu = tf.nn.relu(conv + layer1_biases)\n",
    "                dropout = tf.nn.dropout(relu, keep_rate)\n",
    "                hidden = tf.nn.max_pool(dropout, [1,2,2,1], [1,2,2,1], padding='SAME')\n",
    "                conv = tf.nn.conv2d(hidden, layer2_weights, [1, 1, 1, 1], padding='SAME')\n",
    "                relu = tf.nn.relu(conv + layer2_biases)\n",
    "                dropout = tf.nn.dropout(relu, keep_rate)\n",
    "                hidden = tf.nn.max_pool(dropout, [1,2,2,1], [1,2,2,1], padding='SAME')\n",
    "                shape = hidden.get_shape().as_list()\n",
    "                reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "                hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "                return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
    "              \n",
    "              # Training computation.\n",
    "              logits = model(tf_train_dataset)\n",
    "              loss = tf.reduce_mean(\n",
    "                tf.nn.softmax_cross_entropy_with_logits(logits, tf_train_labels))\n",
    "                \n",
    "              # Optimizer with learning rate decay\n",
    "              global_step = tf.Variable(0)  # count the number of steps taken.\n",
    "              decay_steps = 50\n",
    "              decay_rate = decay_rate_value\n",
    "              learning_rate = tf.train.exponential_decay(initial_learning_rate, global_step, decay_steps, decay_rate)\n",
    "              optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)\n",
    "\n",
    "              \n",
    "              # Predictions for the training, validation, and test data.\n",
    "              train_prediction = tf.nn.softmax(logits)\n",
    "              valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "              test_prediction = tf.nn.softmax(model(tf_test_dataset))\n",
    "\n",
    "\n",
    "            with tf.Session(graph=graph) as session:\n",
    "              tf.initialize_all_variables().run()\n",
    "              print('Initialized kr: {:g}, ilr: {:g}, drv: {:g}'.format(keep_rate, initial_learning_rate, decay_rate_value))\n",
    "              for step in range(num_steps):\n",
    "                offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "                batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "                batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "                feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "                _, l, predictions = session.run(\n",
    "                  [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "                # if (step % 50 == 0):\n",
    "                #   print('Minibatch loss at step %d: %f' % (step, l))\n",
    "                #   print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "                #   print('Validation accuracy: %.1f%%' % accuracy(\n",
    "                #     valid_prediction.eval(), valid_labels))\n",
    "              print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))\n",
    "\n",
    "            "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "Problem 2 - initial grid search\n",
    "---------\n",
    "\n",
    "Best results above\n",
    "\n",
    "__KR__ = keep rate for `dropout()`\n",
    "\n",
    "__ILR__ = initial learning rate\n",
    "\n",
    "__DR__ = learning decay rate percentage per decay steps (`decay_steps=50`)\n",
    "\n",
    "\n",
    "| KR | ILR | DR | Accuracy         \n",
    "| :- |:- | :- | :-:\n",
    "| 0.95 | 0.2 | 0.9 | 91.0%\n",
    "| 0.95 | 0.2 | 0.95 | 91.2%\n",
    "| 0.95 | 0.1 | 0.9 | 90.4%\n",
    "| 0.95 | 0.1 | 0.95 | 91.2%\n",
    "| 0.875 | 0.1 | 0.9 | 90.5%\n",
    "| 0.875 | 0.1 | 0.95 | 91.0%\n",
    "| 0.875 | 0.05 | 0.95 | 90.1%\n",
    "| 0.75 | 0.1 | 0.9 | 90.9%\n",
    "\n",
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized kr: 1, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 89.9%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 91.7%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 91.3%\n",
      "Initialized kr: 1, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 85.1%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 91.5%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 91.2%\n",
      "Initialized kr: 1, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 91.5%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 91.1%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 91.1%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 89.8%\n",
      "Initialized kr: 1, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 90.7%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 1, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 1, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 1, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.3%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.6%\n",
      "Initialized kr: 1, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 87.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 91.9%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 17.2%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 91.4%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 89.9%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 90.9%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 91.6%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 0.95, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 9.9%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 90.9%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 91.2%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 91.1%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 0.95, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 90.9%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 89.8%\n",
      "Initialized kr: 0.95, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 90.1%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 89.5%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.95, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.95, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.95, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 89.9%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.9, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 14.0%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 91.6%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 91.3%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 91.1%\n",
      "Initialized kr: 0.9, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 91.2%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 90.7%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.9, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.9, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.9, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.9, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 86.8%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 91.2%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.85, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 12.0%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 90.9%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 91.8%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 90.7%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 90.9%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 90.7%\n",
      "Initialized kr: 0.85, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 90.1%\n",
      "Initialized kr: 0.85, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.85, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 87.6%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.85, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.1%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 87.3%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.1%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.85, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 86.7%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 83.8%\n",
      "Initialized kr: 0.8, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 90.7%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 91.0%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.8, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 89.8%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 89.8%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.8, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.8, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.7%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 87.1%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.7%\n",
      "Initialized kr: 0.8, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.2%\n",
      "Initialized kr: 0.75, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 90.1%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 89.8%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 90.7%\n",
      "Initialized kr: 0.75, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 90.1%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 91.3%\n",
      "Initialized kr: 0.75, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 90.1%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 91.4%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.75, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.75, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.75, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.3%\n",
      "Initialized kr: 0.75, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 86.4%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.7, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 75.1%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.7, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 46.5%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 86.0%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 90.4%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 90.5%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.7, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 86.8%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 90.6%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.7, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.7, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 89.2%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.7, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.7%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 88.5%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.1%\n",
      "Initialized kr: 0.7, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 87.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 87.2%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 90.3%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.65, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 89.7%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 90.8%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 89.0%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 89.5%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 89.5%\n",
      "Initialized kr: 0.65, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 87.3%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 89.8%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 89.5%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.65, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 87.1%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.65, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 87.6%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 87.4%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.8%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 87.2%\n",
      "Initialized kr: 0.65, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 87.2%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.35, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.3, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.25, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.925\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.9\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.85\n",
      "Test accuracy: 89.1%\n",
      "Initialized kr: 0.6, ilr: 0.2, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.975\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.95\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.925\n",
      "Test accuracy: 90.0%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.9\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.875\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.85\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.15, drv: 0.825\n",
      "Test accuracy: 10.0%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.975\n",
      "Test accuracy: 90.2%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.95\n",
      "Test accuracy: 89.6%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.925\n",
      "Test accuracy: 88.8%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.9\n",
      "Test accuracy: 89.3%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.875\n",
      "Test accuracy: 79.9%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.85\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.6, ilr: 0.1, drv: 0.825\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.975\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.95\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.925\n",
      "Test accuracy: 89.4%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.9\n",
      "Test accuracy: 88.6%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.875\n",
      "Test accuracy: 88.9%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.85\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.6, ilr: 0.05, drv: 0.825\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.975\n",
      "Test accuracy: 88.0%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.95\n",
      "Test accuracy: 88.7%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.925\n",
      "Test accuracy: 88.2%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.9\n",
      "Test accuracy: 87.7%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.875\n",
      "Test accuracy: 87.9%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.85\n",
      "Test accuracy: 88.1%\n",
      "Initialized kr: 0.6, ilr: 0.04, drv: 0.825\n",
      "Test accuracy: 87.2%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.975\n",
      "Test accuracy: 88.3%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.95\n",
      "Test accuracy: 88.4%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.925\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.9\n",
      "Test accuracy: 87.0%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.875\n",
      "Test accuracy: 87.5%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.85\n",
      "Test accuracy: 86.8%\n",
      "Initialized kr: 0.6, ilr: 0.03, drv: 0.825\n",
      "Test accuracy: 86.2%\n"
     ]
    }
   ],
   "source": [
    "for keep_rate in [1, 0.95, 0.9, 0.85, 0.8, 0.75, 0.7, 0.65, 0.6]:\n",
    "    for initial_learning_rate in [0.35, 0.3, 0.25, 0.2, 0.15, 0.1, 0.05, 0.04, 0.03]:\n",
    "        for decay_rate_value in [0.975, 0.95, 0.925, 0.9, 0.875, 0.85, 0.825]:\n",
    "\n",
    "            graph = tf.Graph()\n",
    "\n",
    "            with graph.as_default():\n",
    "\n",
    "              # Input data.\n",
    "              tf_train_dataset = tf.placeholder(\n",
    "                tf.float32, shape=(batch_size, image_size, image_size, num_channels))\n",
    "              tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n",
    "              tf_valid_dataset = tf.constant(valid_dataset)\n",
    "              tf_test_dataset = tf.constant(test_dataset)\n",
    "              \n",
    "              # Variables.\n",
    "              layer1_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [patch_size, patch_size, num_channels, depth], stddev=0.1))\n",
    "              layer1_biases = tf.Variable(tf.zeros([depth]))\n",
    "              layer2_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [patch_size, patch_size, depth, depth], stddev=0.1))\n",
    "              layer2_biases = tf.Variable(tf.constant(1.0, shape=[depth]))\n",
    "              layer3_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [image_size // 4 * image_size // 4 * depth, num_hidden], stddev=0.1))\n",
    "              layer3_biases = tf.Variable(tf.constant(1.0, shape=[num_hidden]))\n",
    "              layer4_weights = tf.Variable(tf.truncated_normal(\n",
    "                  [num_hidden, num_labels], stddev=0.1))\n",
    "              layer4_biases = tf.Variable(tf.constant(1.0, shape=[num_labels]))\n",
    "              \n",
    "              # Model.\n",
    "              # try 1, 2 more layers\n",
    "              def model(data):\n",
    "                conv = tf.nn.conv2d(data, layer1_weights, [1, 1, 1, 1], padding='SAME')\n",
    "                relu = tf.nn.relu(conv + layer1_biases)\n",
    "                dropout = tf.nn.dropout(relu, keep_rate)\n",
    "                hidden = tf.nn.max_pool(dropout, [1,2,2,1], [1,2,2,1], padding='SAME')\n",
    "                conv = tf.nn.conv2d(hidden, layer2_weights, [1, 1, 1, 1], padding='SAME')\n",
    "                relu = tf.nn.relu(conv + layer2_biases)\n",
    "                dropout = tf.nn.dropout(relu, keep_rate)\n",
    "                hidden = tf.nn.max_pool(dropout, [1,2,2,1], [1,2,2,1], padding='SAME')\n",
    "                shape = hidden.get_shape().as_list()\n",
    "                reshape = tf.reshape(hidden, [shape[0], shape[1] * shape[2] * shape[3]])\n",
    "                hidden = tf.nn.relu(tf.matmul(reshape, layer3_weights) + layer3_biases)\n",
    "                return tf.matmul(hidden, layer4_weights) + layer4_biases\n",
    "              \n",
    "              # Training computation.\n",
    "              logits = model(tf_train_dataset)\n",
    "              loss = tf.reduce_mean(\n",
    "                tf.nn.softmax_cross_entropy_with_logits(logits, tf_train_labels))\n",
    "                \n",
    "              # Optimizer with learning rate decay\n",
    "              global_step = tf.Variable(0)  # count the number of steps taken.\n",
    "              decay_steps = 50\n",
    "              decay_rate = decay_rate_value\n",
    "              learning_rate = tf.train.exponential_decay(initial_learning_rate, global_step, decay_steps, decay_rate)\n",
    "              optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)\n",
    "\n",
    "              \n",
    "              # Predictions for the training, validation, and test data.\n",
    "              train_prediction = tf.nn.softmax(logits)\n",
    "              valid_prediction = tf.nn.softmax(model(tf_valid_dataset))\n",
    "              test_prediction = tf.nn.softmax(model(tf_test_dataset))\n",
    "\n",
    "\n",
    "            with tf.Session(graph=graph) as session:\n",
    "              tf.initialize_all_variables().run()\n",
    "              print('Initialized kr: {:g}, ilr: {:g}, drv: {:g}'.format(keep_rate, initial_learning_rate, decay_rate_value))\n",
    "              for step in range(num_steps):\n",
    "                offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n",
    "                batch_data = train_dataset[offset:(offset + batch_size), :, :, :]\n",
    "                batch_labels = train_labels[offset:(offset + batch_size), :]\n",
    "                feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n",
    "                _, l, predictions = session.run(\n",
    "                  [optimizer, loss, train_prediction], feed_dict=feed_dict)\n",
    "                # if (step % 50 == 0):\n",
    "                #   print('Minibatch loss at step %d: %f' % (step, l))\n",
    "                #   print('Minibatch accuracy: %.1f%%' % accuracy(predictions, batch_labels))\n",
    "                #   print('Validation accuracy: %.1f%%' % accuracy(\n",
    "                #     valid_prediction.eval(), valid_labels))\n",
    "              print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))\n",
    "            "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "Problem 2 - second grid search\n",
    "---------\n",
    "\n",
    "__KR__ = keep rate for `dropout()`\n",
    "\n",
    "__ILR__ = initial learning rate\n",
    "\n",
    "__DR__ = learning decay rate percentage per decay steps (`decay_steps=50`)\n",
    "\n",
    "Best result\n",
    "\n",
    "\n",
    "| KR | ILR | DR | Accuracy         \n",
    "| :- |:- | :- | :-:\n",
    "| 0.95 | 0.3 | 0.9 | 91.9%\n",
    "\n",
    "Other good results at or over 91%\n",
    "\n",
    "| KR | ILR | DR | Accuracy         \n",
    "| :- |:- | :- | :-:\n",
    "| 1 | 0.25 | 0.875 | 91.7%\n",
    "| 1 | 0.25 | 0.85 | 91.3%\n",
    "| 1 | 0.2 | 0.925 | 91.0%\n",
    "| 1 | 0.2 | 0.875 | 91.5%\n",
    "| 1 | 0.2 | 0.85 | 91.2%\n",
    "| 1 | 0.15 | 0.975 | 91.5%\n",
    "| 1 | 0.15 | 0.95 | 91.1%\n",
    "| 1 | 0.15 | 0.9 | 91.1%\n",
    "| 1 | 0.15 | 0.875 | 91.0%\n",
    "| 0.95 | 0.25 | 0.95 | 91.4%\n",
    "| 0.95 | 0.2 | 0.875 | 91.6%\n",
    "| 0.95 | 0.15 | 0.95 | 91.2%\n",
    "| 0.95 | 0.15 | 0.875 | 91.1%\n",
    "| 0.95 | 0.1 | 0.95 | 91.0%\n",
    "| 0.95 | 0.1 | 0.925 | 91.0%\n",
    "| 0.9 | 0.15 | 0.975 | 91.0%\n",
    "| 0.9 | 0.15 | 0.95 | 91.6%\n",
    "| 0.9 | 0.15 | 0.925 | 91.3%\n",
    "| 0.9 | 0.15 | 0.85 | 91.1%\n",
    "| 0.9 | 0.1 | 0.95 | 91.2%\n",
    "| 0.9 | 0.1 | 0.9 | 91.0%\n",
    "| 0.85 | 0.2 | 0.9 | 91.2%\n",
    "| 0.85 | 0.15 | 0.925 | 91.8%\n",
    "| 0.85 | 0.1 | 0.95 | 91.0%\n",
    "| 0.8 | 0.15 | 0.975 | 91.0%\n",
    "| 0.8 | 0.1 | 0.95 | 91.0%\n",
    "| 0.75 | 0.15 | 0.85 | 91.3%\n",
    "| 0.75 | 0.1 | 0.925 | 91.4%\n",
    "\n",
    "\n",
    "---"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "default_view": {},
   "name": "4_convolutions.ipynb",
   "provenance": [],
   "version": "0.3.2",
   "views": {}
  },
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}