Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
A mix of autoencoder and a classifier with Tensorflow
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# An autoencoder with classification for MNIST\n",
"Or\n",
"\n",
"## How to create an autoencoder and monitor its embedded space via a classifier\n",
"\n",
"This question was raised on Stack Overflow some weeks ago for *Keras*, and I thought it was a good way to reuse elements from [my recent book](http://blog.audio-tk.com/2018/09/04/book-building-machine-learning-systems-withpython-third-edition/).\n",
"\n",
"Mainly, we will use the MNIST dataset to train an autoencoder, and as a second step we will see how we can add on top of it a classifier that can give us some information on the quality of the embedded space for classification purposes.\n",
"\n",
"So what we will reuse is:\n",
"* the autoencoder concept\n",
"* part of the discriminator and the generator from our GAN\n",
"* the MNIST classifier final layers\n",
"* Tensorboard usage"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Retrieving the data\n",
"\n",
"Let's start with some common code for the two steps."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import tensorflow as tf\n",
"import numpy as np\n",
"from matplotlib import pyplot as plt\n",
"from matplotlib import offsetbox\n",
"from sklearn.metrics import confusion_matrix\n",
"\n",
"%matplotlib inline\n",
"\n",
"n_epochs = 10\n",
"learning_rate = 0.0002\n",
"batch_size = 128\n",
"image_shape = [28,28,1]\n",
"step = batch_size * 100\n",
"dim_W1 = 128\n",
"dim_W2 = 64\n",
"dim_W3 = 32\n",
"dim_C1 = 16\n",
"\n",
"dim_embedded = 2"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"def plot_confusion_matrix(cm, genre_list, title):\n",
" plt.figure(num=None, figsize=(5, 4))\n",
" ax = plt.axes()\n",
" im = ax.matshow(cm, cmap='Blues', vmin=0, vmax=1.0)\n",
" ax.set_xticks(range(len(genre_list)))\n",
" ax.set_xticklabels(genre_list)\n",
" ax.xaxis.set_ticks_position(\"bottom\")\n",
" ax.set_yticks(range(len(genre_list)))\n",
" ax.set_yticklabels(genre_list)\n",
" ax.tick_params(axis='both', which='both', bottom=False, left=False)\n",
" plt.title(title)\n",
" plt.colorbar(im, ax=ax)\n",
" plt.grid(False)\n",
" plt.xlabel('Predicted class')\n",
" plt.ylabel('True class')\n",
" \n",
"def plot_embedding(X, labels, title=None):\n",
" x_min, x_max = np.min(X, 0), np.max(X, 0)\n",
" X = (X - x_min) / (x_max - x_min)\n",
"\n",
" plt.figure()\n",
" ax = plt.subplot(111)\n",
" for i in range(X.shape[0]):\n",
" plt.text(X[i, 0], X[i, 1], str(labels[i]),\n",
" color=plt.cm.Set1(labels[i] / 10.),\n",
" fontdict={'weight': 'bold', 'size': 9})\n",
" plt.xticks([]), plt.yticks([])\n",
" if title is not None:\n",
" plt.title(title)"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"WARNING:tensorflow:From <ipython-input-3-2aed5ca29197>:2: read_data_sets (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
"Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n",
"WARNING:tensorflow:From /home/matthieu/miniconda3/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:260: maybe_download (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
"Please write your own downloading logic.\n",
"WARNING:tensorflow:From /home/matthieu/miniconda3/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:262: extract_images (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
"Please use tf.data to implement this functionality.\n",
"Extracting MNIST_data/train-images-idx3-ubyte.gz\n",
"WARNING:tensorflow:From /home/matthieu/miniconda3/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:267: extract_labels (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
"Please use tf.data to implement this functionality.\n",
"Extracting MNIST_data/train-labels-idx1-ubyte.gz\n",
"Extracting MNIST_data/t10k-images-idx3-ubyte.gz\n",
"Extracting MNIST_data/t10k-labels-idx1-ubyte.gz\n",
"WARNING:tensorflow:From /home/matthieu/miniconda3/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py:290: DataSet.__init__ (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
"Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n"
]
}
],
"source": [
"from tensorflow.examples.tutorials.mnist import input_data\n",
"mnist = input_data.read_data_sets(\"MNIST_data/\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The future-proof way of getting MNIST for tensorflow can be found at https://github.com/tensorflow/models/blob/master/official/mnist/dataset.py\n",
"\n",
"We know reshape the images as we need:"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"mnist.train.images.shape = -1, 28, 28, 1\n",
"mnist.test.images.shape = -1, 28, 28, 1\n",
"\n",
"num_train = mnist.train.images.shape[0]\n",
"num_test = mnist.test.images.shape[0]"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"def batchnormalize(X, eps=1e-8, g=None, b=None):\n",
" if X.get_shape().ndims == 4:\n",
" mean = tf.reduce_mean(X, [0,1,2])\n",
" std = tf.reduce_mean( tf.square(X-mean), [0,1,2] )\n",
" X = (X-mean) / tf.sqrt(std+eps)\n",
"\n",
" if g is not None and b is not None:\n",
" g = tf.reshape(g, [1,1,1,-1])\n",
" b = tf.reshape(b, [1,1,1,-1])\n",
" X = X*g + b\n",
"\n",
" elif X.get_shape().ndims == 2:\n",
" mean = tf.reduce_mean(X, 0)\n",
" std = tf.reduce_mean(tf.square(X-mean), 0)\n",
" X = (X-mean) / tf.sqrt(std+eps)\n",
"\n",
" if g is not None and b is not None:\n",
" g = tf.reshape(g, [1,-1])\n",
" b = tf.reshape(b, [1,-1])\n",
" X = X*g + b\n",
"\n",
" else:\n",
" raise NotImplementedError\n",
"\n",
" return X\n",
"\n",
"class AutoEncoder():\n",
" def __init__(\n",
" self,\n",
" image_shape=[28,28,1],\n",
" dim_W1=1024,\n",
" dim_W2=128,\n",
" dim_W3=64,\n",
" dim_embedded=2,\n",
" ):\n",
"\n",
" self.image_shape = image_shape\n",
"\n",
" self.dim_W1 = dim_W1\n",
" self.dim_W2 = dim_W2\n",
" self.dim_W3 = dim_W3\n",
" self.dim_embedded = dim_embedded\n",
"\n",
" def build_model(self):\n",
"\n",
" image = tf.placeholder(tf.float32, [None]+self.image_shape)\n",
" embedded = self.encode(image)\n",
" decoded = self.decode(embedded)\n",
" \n",
" # We clip the output, as 0 and 1 cannot be achieved with a sigmoid output\n",
" logits = tf.clip_by_value(decoded, 1e-7, 1. - 1e-7)\n",
" \n",
" cost_autoencoder = tf.reduce_mean(tf.square(logits - image))\n",
"\n",
" summaries = tf.summary.merge([\n",
" tf.summary.scalar(\"loss/train\", cost_autoencoder),\n",
" ])\n",
" summaries_test = tf.summary.merge([\n",
" tf.summary.scalar(\"loss/test\", cost_autoencoder),\n",
" ])\n",
"\n",
" return image, embedded, decoded, cost_autoencoder, summaries, summaries_test\n",
"\n",
" def create_conv2d(self, input, filters, kernel_size, name):\n",
" layer = tf.layers.conv2d(\n",
" inputs=input,\n",
" filters=filters,\n",
" kernel_size=kernel_size,\n",
" strides=[2,2],\n",
" name=\"Conv2d_\" + name,\n",
" padding=\"SAME\")\n",
" layer = tf.nn.leaky_relu(layer, name= \"LeakyRELU\" + name)\n",
" return layer\n",
"\n",
" def create_conv2d_transpose(self, input, filters, kernel_size, name, with_batch_norm):\n",
" layer = tf.layers.conv2d_transpose(\n",
" inputs=input,\n",
" filters=filters,\n",
" kernel_size=kernel_size,\n",
" strides=[2,2],\n",
" name=\"Conv2d_\" + name,\n",
" padding=\"SAME\")\n",
" if with_batch_norm:\n",
" layer = batchnormalize(layer)\n",
" layer = tf.nn.relu(layer)\n",
" return layer\n",
"\n",
" def create_dense(self, input, units, name, leaky):\n",
" layer = tf.layers.dense(\n",
" inputs=input,\n",
" units=units,\n",
" name=\"Dense\" + name,\n",
" )\n",
" layer = batchnormalize(layer)\n",
" if leaky:\n",
" layer = tf.nn.leaky_relu(layer, name= \"LeakyRELU\" + name)\n",
" else:\n",
" layer = tf.nn.relu(layer, name=\"RELU_\" + name)\n",
" return layer\n",
"\n",
" def encode(self, image):\n",
" with tf.variable_scope('encoder'):\n",
" h1 = self.create_conv2d(image, self.dim_W3, 5, \"Layer1\")\n",
" \n",
" h2 = self.create_conv2d(h1, self.dim_W2, 5, \"Layer2\")\n",
" h2 = tf.reshape(h2, tf.stack([-1, 7*7*self.dim_W2]))\n",
" \n",
" h3 = self.create_dense(h2, self.dim_W1, \"Layer3\", True)\n",
" \n",
" h4 = self.create_dense(h3, self.dim_embedded, \"Layer4\", True)\n",
" return h4\n",
"\n",
" def decode(self, embedded):\n",
" with tf.variable_scope('decoder'):\n",
"\n",
" h1 = self.create_dense(embedded, self.dim_W1, \"Layer1\", False)\n",
"\n",
" h2 = self.create_dense(h1, self.dim_W2*7*7, \"Layer2\", False)\n",
" h2 = tf.reshape(h2, tf.stack([-1,7,7,self.dim_W2]))\n",
"\n",
" h3 = self.create_conv2d_transpose(h2, self.dim_W3, 5, \"Layer3\", True)\n",
"\n",
" h4 = self.create_conv2d_transpose(h3, 1, 7, \"Layer4\", False)\n",
" x = tf.nn.sigmoid(h4)\n",
" return x"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Let's get our model now."
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"tf.reset_default_graph()\n",
"autoencoder_model = AutoEncoder(\n",
" image_shape=image_shape,\n",
" dim_W1=dim_W1,\n",
" dim_W2=dim_W2,\n",
" dim_W3=dim_W3,\n",
" dim_embedded=dim_embedded\n",
" )\n",
"image, embedded, decoded, cost_autoencoder, summaries, summaries_test = autoencoder_model.build_model()"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"train_op_autoencoder = tf.train.AdamOptimizer(learning_rate, beta1=0.5).minimize(cost_autoencoder)\n",
"summary_writer = tf.summary.FileWriter(\"/tmp/tensorboard/part1\", tf.get_default_graph())"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 0\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 1\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 2\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 3\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 4\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 5\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 6\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 7\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 8\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 9\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n"
]
}
],
"source": [
"with tf.Session() as sess:\n",
" sess.run(tf.global_variables_initializer())\n",
" for epoch in range(n_epochs):\n",
" permut = np.random.permutation(num_train)\n",
" trX = mnist.train.images[permut]\n",
"\n",
" print(\"epoch: %i\" % epoch)\n",
" for j in range(0, num_train, batch_size):\n",
" if j % step == 0:\n",
" print(\" batch: %i\" % j)\n",
"\n",
" batch = permut[j:j+batch_size]\n",
"\n",
" Xs = trX[batch]\n",
"\n",
" _, local_summaries = sess.run([train_op_autoencoder, summaries],\n",
" feed_dict={\n",
" image:Xs,\n",
" })\n",
" summary_writer.add_summary(local_summaries, epoch * num_train + j)\n",
" local_test_summaries = sess.run(summaries_test,\n",
" feed_dict={\n",
" image:mnist.test.images,\n",
" })\n",
" summary_writer.add_summary(local_test_summaries, epoch * num_train)\n",
"\n",
" embedded_space_train = sess.run(embedded,\n",
" feed_dict={\n",
" image:mnist.train.images,\n",
" })\n",
" embedded_space_test = sess.run(embedded,\n",
" feed_dict={\n",
" image:mnist.test.images,\n",
" })"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plot_embedding(embedded_space_train, mnist.train.labels)"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"plot_embedding(embedded_space_test, mnist.test.labels)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Adding a classification accuracy measure on the fly\n",
"\n",
"Let's modify now the autoencoder to also add a classification layer."
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {},
"outputs": [],
"source": [
"class AutoEncoderClassifier():\n",
" def __init__(\n",
" self,\n",
" image_shape=[28,28,1],\n",
" dim_W1=128,\n",
" dim_W2=64,\n",
" dim_W3=32,\n",
" dim_embedded=2,\n",
" dim_C1=32,\n",
" ):\n",
"\n",
" self.image_shape = image_shape\n",
"\n",
" self.dim_W1 = dim_W1\n",
" self.dim_W2 = dim_W2\n",
" self.dim_W3 = dim_W3\n",
" self.dim_embedded = dim_embedded\n",
" self.dim_C1 = dim_C1\n",
"\n",
" def build_model(self):\n",
"\n",
" image = tf.placeholder(tf.float32, [None]+self.image_shape)\n",
" label = tf.placeholder(tf.int64, [None])\n",
" embedded = self.encode(image)\n",
" decoded = self.decode(embedded)\n",
" classified = self.classify(embedded)\n",
" \n",
" # We clip the output, as 0 and 1 cannot be achieved with a sigmoid output\n",
" logits = tf.clip_by_value(decoded, 1e-7, 1. - 1e-7)\n",
" \n",
" cost_autoencoder = tf.reduce_mean(tf.square(logits - image))\n",
" \n",
" cost_classify = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=label, logits=classified))\n",
" accuracy_classify = tf.reduce_mean(tf.cast(tf.equal(tf.argmax(classified, axis=1), label), tf.float32), name=\"accuracy\")\n",
" \n",
" summaries_train = tf.summary.merge([\n",
" tf.summary.scalar(\"loss/train\", cost_autoencoder),\n",
" ])\n",
" summaries_classifier = tf.summary.merge([\n",
" tf.summary.scalar(\"accuracy/train\", accuracy_classify),\n",
" ])\n",
" summaries_all_test = tf.summary.merge([\n",
" tf.summary.scalar(\"loss/test\", cost_autoencoder),\n",
" tf.summary.scalar(\"accuracy/test\", accuracy_classify),\n",
" ])\n",
" \n",
" return image, label, embedded, decoded, classified, cost_autoencoder, cost_classify, accuracy_classify, summaries_train, summaries_classifier, summaries_all_test\n",
"\n",
" def create_conv2d(self, input, filters, kernel_size, name):\n",
" layer = tf.layers.conv2d(\n",
" inputs=input,\n",
" filters=filters,\n",
" kernel_size=kernel_size,\n",
" strides=[2,2],\n",
" name=\"Conv2d_\" + name,\n",
" padding=\"SAME\")\n",
" layer = tf.nn.leaky_relu(layer, name= \"LeakyRELU\" + name)\n",
" return layer\n",
"\n",
" def create_conv2d_transpose(self, input, filters, kernel_size, name, with_batch_norm):\n",
" layer = tf.layers.conv2d_transpose(\n",
" inputs=input,\n",
" filters=filters,\n",
" kernel_size=kernel_size,\n",
" strides=[2,2],\n",
" name=\"Conv2d_\" + name,\n",
" padding=\"SAME\")\n",
" if with_batch_norm:\n",
" layer = batchnormalize(layer)\n",
" layer = tf.nn.relu(layer)\n",
" return layer\n",
"\n",
" def create_dense(self, input, units, name, leaky):\n",
" layer = tf.layers.dense(\n",
" inputs=input,\n",
" units=units,\n",
" name=\"Dense\" + name,\n",
" )\n",
" layer = batchnormalize(layer)\n",
" if leaky:\n",
" layer = tf.nn.leaky_relu(layer, name= \"LeakyRELU\" + name)\n",
" else:\n",
" layer = tf.nn.relu(layer, name=\"RELU_\" + name)\n",
" return layer\n",
"\n",
" def encode(self, image):\n",
" with tf.variable_scope('encoder'):\n",
" h1 = self.create_conv2d(image, self.dim_W3, 5, \"Layer1\")\n",
" \n",
" h2 = self.create_conv2d(h1, self.dim_W2, 5, \"Layer2\")\n",
" h2 = tf.reshape(h2, tf.stack([-1, 7*7*self.dim_W2]))\n",
" \n",
" h3 = self.create_dense(h2, self.dim_W1, \"Layer3\", True)\n",
" \n",
" h4 = self.create_dense(h3, self.dim_embedded, \"Layer4\", True)\n",
" return h4\n",
"\n",
" def decode(self, embedded):\n",
" with tf.variable_scope('decoder'):\n",
"\n",
" h1 = self.create_dense(embedded, self.dim_W1, \"Layer1\", False)\n",
"\n",
" h2 = self.create_dense(h1, self.dim_W2*7*7, \"Layer2\", False)\n",
" h2 = tf.reshape(h2, tf.stack([-1,7,7,self.dim_W2]))\n",
"\n",
" h3 = self.create_conv2d_transpose(h2, self.dim_W3, 5, \"Layer3\", True)\n",
"\n",
" h4 = self.create_conv2d_transpose(h3, 1, 7, \"Layer4\", False)\n",
" x = tf.nn.sigmoid(h4)\n",
" return x\n",
" \n",
"\n",
" def classify(self, embedded):\n",
" with tf.variable_scope('classifier'):\n",
"\n",
" h1 = self.create_dense(embedded, self.dim_C1, \"Layer1\", False)\n",
"\n",
" h2 = self.create_dense(h1, 10, \"Layer2\", False)\n",
" return h2"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {},
"outputs": [],
"source": [
"tf.reset_default_graph()\n",
"autoencoder_classifier_model = AutoEncoderClassifier(\n",
" image_shape=image_shape,\n",
" dim_W1=dim_W1,\n",
" dim_W2=dim_W2,\n",
" dim_W3=dim_W3,\n",
" dim_embedded=dim_embedded,\n",
" dim_C1=dim_C1,\n",
" )\n",
"\n",
"image, label, embedded, decoded, classified, cost_autoencoder, cost_classify, accuracy_classify, summaries_train, summaries_classifier, summaries_all_test = autoencoder_classifier_model.build_model()"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {},
"outputs": [],
"source": [
"encode_vars = list(filter(lambda x: x.name.startswith('encode'), tf.trainable_variables()))\n",
"decode_vars = list(filter(lambda x: x.name.startswith('decode'), tf.trainable_variables()))\n",
"classify_vars = list(filter(lambda x: x.name.startswith('classifier'), tf.trainable_variables()))\n",
"\n",
"train_op_autoencoder = tf.train.AdamOptimizer(learning_rate, beta1=0.5).minimize(cost_autoencoder, var_list=encode_vars+decode_vars)\n",
"train_op_classify = tf.train.AdamOptimizer(learning_rate, beta1=0.5).minimize(cost_classify, var_list=classify_vars)\n",
"\n",
"summary_writer = tf.summary.FileWriter(\"/tmp/tensorboard/part2\", tf.get_default_graph())"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"As for the GAN case, we now alternate between optimizing the autoencoder and then the classifier."
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 0\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 1\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 2\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 3\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 4\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 5\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 6\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 7\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 8\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 9\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 10\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 11\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 12\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 13\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 14\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 15\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 16\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 17\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 18\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 19\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"Cost for the training set: 0.038948\n",
"Cost for the testing set: 0.039474\n"
]
}
],
"source": [
"with tf.Session() as sess:\n",
" sess.run(tf.global_variables_initializer())\n",
" for epoch in range(2*n_epochs):\n",
" permut = np.random.permutation(num_train)\n",
" trX = mnist.train.images[permut]\n",
" trY = mnist.train.labels[permut]\n",
"\n",
" print(\"epoch: %i\" % epoch)\n",
" for j in range(0, num_train, batch_size):\n",
" if j % step == 0:\n",
" print(\" batch: %i\" % j)\n",
"\n",
" batch = permut[j:j+batch_size]\n",
"\n",
" Xs = trX[batch]\n",
" Ys = trY[batch]\n",
"\n",
" if j % (2 * batch_size) == 0:\n",
" _, local_summaries = sess.run([train_op_autoencoder, summaries_train],\n",
" feed_dict={\n",
" image:Xs,\n",
" })\n",
" summary_writer.add_summary(local_summaries, epoch * num_train + j)\n",
" else:\n",
" _, local_summary_1, local_summary_2 = sess.run([train_op_classify, summaries_train, summaries_classifier],\n",
" feed_dict={\n",
" image:Xs,\n",
" label:Ys,\n",
" })\n",
" summary_writer.add_summary(local_summary_1, epoch * num_train + j)\n",
" summary_writer.add_summary(local_summary_2, epoch * num_train + j)\n",
"\n",
" local_test_summaries = sess.run(summaries_all_test,\n",
" feed_dict={\n",
" image:mnist.test.images,\n",
" label:mnist.test.labels,\n",
" })\n",
" summary_writer.add_summary(local_test_summaries, epoch * num_train)\n",
" \n",
"\n",
" embedded_space_train, cost_train = sess.run([embedded, cost_autoencoder],\n",
" feed_dict={\n",
" image:mnist.train.images,\n",
" })\n",
" embedded_space_test, classify_test, cost_test = sess.run([embedded, classified, cost_autoencoder],\n",
" feed_dict={\n",
" image:mnist.test.images,\n",
" })\n",
"\n",
"print(\"Cost for the training set: %f\" % cost_train)\n",
"print(\"Cost for the testing set: %f\" % cost_test)\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If we display the result, we can see that we get quite a good result for classification in 2D, although the autoencoder loss is quite bad. This is logical considering that 2 dimensions are not enough to represent the complexit of digits!"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 360x288 with 2 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"class_test = np.argmax(classify_test, axis=1)\n",
"cm = confusion_matrix(mnist.test.labels, class_test)\n",
"plot_confusion_matrix(cm / np.sum(cm, axis=0), list(range(10)), \"Autoencoder performance for classification in 2D\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Let's see what happens in 3D"
]
},
{
"cell_type": "code",
"execution_count": 16,
"metadata": {},
"outputs": [],
"source": [
"tf.reset_default_graph()\n",
"autoencoder_classifier_model = AutoEncoderClassifier(\n",
" image_shape=image_shape,\n",
" dim_W1=dim_W1,\n",
" dim_W2=dim_W2,\n",
" dim_W3=dim_W3,\n",
" dim_embedded=3,\n",
" dim_C1=dim_C1,\n",
" )\n",
"\n",
"image, label, embedded, decoded, classified, cost_autoencoder, cost_classify, accuracy_classify, summaries_train, summaries_classifier, summaries_all_test = autoencoder_classifier_model.build_model()"
]
},
{
"cell_type": "code",
"execution_count": 17,
"metadata": {},
"outputs": [],
"source": [
"encode_vars = list(filter(lambda x: x.name.startswith('encode'), tf.trainable_variables()))\n",
"decode_vars = list(filter(lambda x: x.name.startswith('decode'), tf.trainable_variables()))\n",
"classify_vars = list(filter(lambda x: x.name.startswith('classifier'), tf.trainable_variables()))\n",
"\n",
"train_op_autoencoder = tf.train.AdamOptimizer(learning_rate, beta1=0.5).minimize(cost_autoencoder, var_list=encode_vars+decode_vars)\n",
"train_op_classify = tf.train.AdamOptimizer(learning_rate, beta1=0.5).minimize(cost_classify, var_list=classify_vars)\n",
"\n",
"summary_writer = tf.summary.FileWriter(\"/tmp/tensorboard/part3\", tf.get_default_graph())"
]
},
{
"cell_type": "code",
"execution_count": 18,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"epoch: 0\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 1\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 2\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 3\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 4\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 5\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 6\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 7\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 8\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 9\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 10\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 11\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 12\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 13\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 14\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 15\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 16\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 17\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 18\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"epoch: 19\n",
" batch: 0\n",
" batch: 12800\n",
" batch: 25600\n",
" batch: 38400\n",
" batch: 51200\n",
"Cost for the training set: 0.032497\n",
"Cost for the testing set: 0.033085\n"
]
}
],
"source": [
"with tf.Session() as sess:\n",
" sess.run(tf.global_variables_initializer())\n",
" for epoch in range(2*n_epochs):\n",
" permut = np.random.permutation(num_train)\n",
" trX = mnist.train.images[permut]\n",
" trY = mnist.train.labels[permut]\n",
"\n",
" print(\"epoch: %i\" % epoch)\n",
" for j in range(0, num_train, batch_size):\n",
" if j % step == 0:\n",
" print(\" batch: %i\" % j)\n",
"\n",
" batch = permut[j:j+batch_size]\n",
"\n",
" Xs = trX[batch]\n",
" Ys = trY[batch]\n",
"\n",
" if j % (2 * batch_size) == 0:\n",
" _, local_summaries = sess.run([train_op_autoencoder, summaries_train],\n",
" feed_dict={\n",
" image:Xs,\n",
" })\n",
" summary_writer.add_summary(local_summaries, epoch * num_train + j)\n",
" else:\n",
" _, local_summary_1, local_summary_2 = sess.run([train_op_classify, summaries_train, summaries_classifier],\n",
" feed_dict={\n",
" image:Xs,\n",
" label:Ys,\n",
" })\n",
" summary_writer.add_summary(local_summary_1, epoch * num_train + j)\n",
" summary_writer.add_summary(local_summary_2, epoch * num_train + j)\n",
"\n",
" local_test_summaries = sess.run(summaries_all_test,\n",
" feed_dict={\n",
" image:mnist.test.images,\n",
" label:mnist.test.labels,\n",
" })\n",
" summary_writer.add_summary(local_test_summaries, epoch * num_train)\n",
"\n",
" embedded_space_train, cost_train = sess.run([embedded, cost_autoencoder],\n",
" feed_dict={\n",
" image:mnist.train.images,\n",
" })\n",
" embedded_space_test, classify_test, cost_test = sess.run([embedded, classified, cost_autoencoder],\n",
" feed_dict={\n",
" image:mnist.test.images,\n",
" })\n",
"\n",
"print(\"Cost for the training set: %f\" % cost_train)\n",
"print(\"Cost for the testing set: %f\" % cost_test)\n"
]
},
{
"cell_type": "code",
"execution_count": 19,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 360x288 with 2 Axes>"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"class_test = np.argmax(classify_test, axis=1)\n",
"cm = confusion_matrix(mnist.test.labels, class_test)\n",
"plot_confusion_matrix(cm / np.sum(cm, axis=0), list(range(10)), \"Autoencoder performance for classification in 3D\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.5"
}
},
"nbformat": 4,
"nbformat_minor": 2
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.