Learning Rate.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import the libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.callbacks import Callback\n",
    "import keras.backend as K\n",
    "\n",
    "import keras\n",
    "from keras.datasets import mnist\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from lr_utils.keras_lr_finder import LRFinder\n",
    "from lr_utils.keras_SGDR import SGDRScheduler\n",
    "from lr_utils.keras_CLR import CyclicLR\n",
    "from keras.callbacks import EarlyStopping"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "60000 train samples\n",
      "10000 test samples\n"
     ]
    }
   ],
   "source": [
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
    "\n",
    "X_train = X_train.reshape(60000, 784)\n",
    "X_test = X_test.reshape(10000, 784)\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "X_train /= 255\n",
    "X_test /= 255\n",
    "print(X_train.shape[0], 'train samples')\n",
    "print(X_test.shape[0], 'test samples')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_classes = 10\n",
    "batch_size = 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create a Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The loss function has to be **categorical_crossentropy** since we deal with categorical labels.\n",
    "The same goes for the last activation which has to be **softmax** since we have more than two classes (sigmoid would be fine for two classes)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_4 (Dense)              (None, 512)               401920    \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 669,706\n",
      "Trainable params: 669,706\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(512, activation='relu', input_shape=(784,)))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(512, activation='relu'))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(num_classes, activation='softmax'))\n",
    "\n",
    "model.summary()\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"410pt\" viewBox=\"0.00 0.00 175.24 410.00\" width=\"175pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(1 1) rotate(0) translate(4 406)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"#ffffff\" points=\"-4,4 -4,-406 171.2392,-406 171.2392,4 -4,4\" stroke=\"transparent\"/>\n",
       "<!-- 4813760832 -->\n",
       "<g class=\"node\" id=\"node1\">\n",
       "<title>4813760832</title>\n",
       "<polygon fill=\"none\" points=\"0,-365.5 0,-401.5 167.2392,-401.5 167.2392,-365.5 0,-365.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.6196\" y=\"-379.3\">dense_4_input: InputLayer</text>\n",
       "</g>\n",
       "<!-- 4813760272 -->\n",
       "<g class=\"node\" id=\"node2\">\n",
       "<title>4813760272</title>\n",
       "<polygon fill=\"none\" points=\"31.4972,-292.5 31.4972,-328.5 135.742,-328.5 135.742,-292.5 31.4972,-292.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.6196\" y=\"-306.3\">dense_4: Dense</text>\n",
       "</g>\n",
       "<!-- 4813760832&#45;&gt;4813760272 -->\n",
       "<g class=\"edge\" id=\"edge1\">\n",
       "<title>4813760832-&gt;4813760272</title>\n",
       "<path d=\"M83.6196,-365.4551C83.6196,-357.3828 83.6196,-347.6764 83.6196,-338.6817\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"87.1197,-338.5903 83.6196,-328.5904 80.1197,-338.5904 87.1197,-338.5903\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "<!-- 4813759096 -->\n",
       "<g class=\"node\" id=\"node3\">\n",
       "<title>4813759096</title>\n",
       "<polygon fill=\"none\" points=\"19.8198,-219.5 19.8198,-255.5 147.4194,-255.5 147.4194,-219.5 19.8198,-219.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.6196\" y=\"-233.3\">dropout_3: Dropout</text>\n",
       "</g>\n",
       "<!-- 4813760272&#45;&gt;4813759096 -->\n",
       "<g class=\"edge\" id=\"edge2\">\n",
       "<title>4813760272-&gt;4813759096</title>\n",
       "<path d=\"M83.6196,-292.4551C83.6196,-284.3828 83.6196,-274.6764 83.6196,-265.6817\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"87.1197,-265.5903 83.6196,-255.5904 80.1197,-265.5904 87.1197,-265.5903\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "<!-- 4813759824 -->\n",
       "<g class=\"node\" id=\"node4\">\n",
       "<title>4813759824</title>\n",
       "<polygon fill=\"none\" points=\"31.4972,-146.5 31.4972,-182.5 135.742,-182.5 135.742,-146.5 31.4972,-146.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.6196\" y=\"-160.3\">dense_5: Dense</text>\n",
       "</g>\n",
       "<!-- 4813759096&#45;&gt;4813759824 -->\n",
       "<g class=\"edge\" id=\"edge3\">\n",
       "<title>4813759096-&gt;4813759824</title>\n",
       "<path d=\"M83.6196,-219.4551C83.6196,-211.3828 83.6196,-201.6764 83.6196,-192.6817\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"87.1197,-192.5903 83.6196,-182.5904 80.1197,-192.5904 87.1197,-192.5903\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "<!-- 4813758536 -->\n",
       "<g class=\"node\" id=\"node5\">\n",
       "<title>4813758536</title>\n",
       "<polygon fill=\"none\" points=\"19.8198,-73.5 19.8198,-109.5 147.4194,-109.5 147.4194,-73.5 19.8198,-73.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.6196\" y=\"-87.3\">dropout_4: Dropout</text>\n",
       "</g>\n",
       "<!-- 4813759824&#45;&gt;4813758536 -->\n",
       "<g class=\"edge\" id=\"edge4\">\n",
       "<title>4813759824-&gt;4813758536</title>\n",
       "<path d=\"M83.6196,-146.4551C83.6196,-138.3828 83.6196,-128.6764 83.6196,-119.6817\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"87.1197,-119.5903 83.6196,-109.5904 80.1197,-119.5904 87.1197,-119.5903\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "<!-- 4813708872 -->\n",
       "<g class=\"node\" id=\"node6\">\n",
       "<title>4813708872</title>\n",
       "<polygon fill=\"none\" points=\"31.4972,-.5 31.4972,-36.5 135.742,-36.5 135.742,-.5 31.4972,-.5\" stroke=\"#000000\"/>\n",
       "<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"83.6196\" y=\"-14.3\">dense_6: Dense</text>\n",
       "</g>\n",
       "<!-- 4813758536&#45;&gt;4813708872 -->\n",
       "<g class=\"edge\" id=\"edge5\">\n",
       "<title>4813758536-&gt;4813708872</title>\n",
       "<path d=\"M83.6196,-73.4551C83.6196,-65.3828 83.6196,-55.6764 83.6196,-46.6817\" fill=\"none\" stroke=\"#000000\"/>\n",
       "<polygon fill=\"#000000\" points=\"87.1197,-46.5903 83.6196,-36.5904 80.1197,-46.5904 87.1197,-46.5903\" stroke=\"#000000\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import SVG\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "\n",
    "SVG(model_to_dot(model).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Find the optimal Learning Rate"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This method has been proposed by Leslie Smith. The concept to find the best learning rate is to begin at a small one, and increase it progressively. By looking at the loss function for the values of the Learning Rate, we can see it being very stable (no improvement) when the LR is small, then drop faster and faster and finally oscillate and diverge when the LR is too big. The optimal LR stands in the middle, where the slope is the steepest (not at the minimum of the curve)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/1\n",
      "38496/60000 [==================>...........] - ETA: 9s - loss: 0.9724 - acc: 0.7751"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x11eeb8358>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr_finder = LRFinder(min_lr=1e-5, max_lr=1e-2, steps_per_epoch=np.ceil(X_train.shape[0]/batch_size), epochs=1)\n",
    "model.fit(X_train, y_train, callbacks=[lr_finder] )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Raw loss values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11eeb8518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr_finder.plot_loss()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Smoothed loss values. It is now easier to find the optimal learning rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11f53d1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr_finder.plot_avg_loss()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Stochastic Gradient Descent with Restart (SGDR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 10s 165us/step - loss: 0.4421 - acc: 0.9030 - val_loss: 0.2467 - val_acc: 0.9422\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 9s 155us/step - loss: 0.3167 - acc: 0.9261 - val_loss: 0.2216 - val_acc: 0.9494\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.2895 - acc: 0.9305 - val_loss: 0.2118 - val_acc: 0.9511\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.2686 - acc: 0.9351 - val_loss: 0.1835 - val_acc: 0.9555\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.2250 - acc: 0.9425 - val_loss: 0.1668 - val_acc: 0.9605\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.2032 - acc: 0.9482 - val_loss: 0.1573 - val_acc: 0.9637\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 9s 156us/step - loss: 0.1865 - acc: 0.9500 - val_loss: 0.1514 - val_acc: 0.9640\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 10s 158us/step - loss: 0.1776 - acc: 0.9529 - val_loss: 0.1490 - val_acc: 0.9649\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1749 - acc: 0.9522 - val_loss: 0.1383 - val_acc: 0.9668\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 9s 156us/step - loss: 0.1550 - acc: 0.9577 - val_loss: 0.1304 - val_acc: 0.9674\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1241ac8d0>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "schedule = SGDRScheduler(min_lr=1e-4,max_lr=1e-3,steps_per_epoch=np.ceil(X_train.shape[0]/batch_size), lr_decay=0.9, cycle_length=3,mult_factor=1.5)\n",
    "model.fit(X_train, y_train, epochs=10, batch_size=128, validation_data=(X_test, y_test), callbacks=[schedule])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1241ac1d0>]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1241acb38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"SGDR\")\n",
    "plt.plot(schedule.history['iterations'], schedule.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.149040305674\n",
      "Test accuracy: 0.9649\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(X_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Cyclical Learning Rate (CLR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular = CyclicLR(mode='triangular', step_size=np.ceil(X_train.shape[0]/batch_size))\n",
    "\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.2283 - acc: 0.9437 - val_loss: 0.2168 - val_acc: 0.9527\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 9s 155us/step - loss: 0.2321 - acc: 0.9442 - val_loss: 0.1250 - val_acc: 0.9701\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1699 - acc: 0.9563 - val_loss: 0.1628 - val_acc: 0.9583\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.1809 - acc: 0.9544 - val_loss: 0.1105 - val_acc: 0.9729\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1408 - acc: 0.9637 - val_loss: 0.1570 - val_acc: 0.9627\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1701 - acc: 0.9575 - val_loss: 0.1161 - val_acc: 0.9724\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1312 - acc: 0.9660 - val_loss: 0.1543 - val_acc: 0.9673\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.1622 - acc: 0.9616 - val_loss: 0.1128 - val_acc: 0.9744\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1207 - acc: 0.9688 - val_loss: 0.1473 - val_acc: 0.9685\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1567 - acc: 0.9627 - val_loss: 0.1133 - val_acc: 0.9762\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1241a0400>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data=(X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1254c6518>]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1254c65c0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"CLR - 'triangular' Policy\")\n",
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular = CyclicLR(mode='triangular2', step_size=np.ceil(X_train.shape[0]/batch_size))\n",
    "\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.1353 - acc: 0.9678 - val_loss: 0.1668 - val_acc: 0.9626\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.1595 - acc: 0.9635 - val_loss: 0.1201 - val_acc: 0.9736\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.1010 - acc: 0.9746 - val_loss: 0.1361 - val_acc: 0.9702\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0956 - acc: 0.9761 - val_loss: 0.1023 - val_acc: 0.9776\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0746 - acc: 0.9808 - val_loss: 0.1086 - val_acc: 0.9761\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0706 - acc: 0.9818 - val_loss: 0.1063 - val_acc: 0.9784\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0616 - acc: 0.9835 - val_loss: 0.1078 - val_acc: 0.9780\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0619 - acc: 0.9839 - val_loss: 0.1017 - val_acc: 0.9784\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0552 - acc: 0.9852 - val_loss: 0.1029 - val_acc: 0.9789\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0563 - acc: 0.9845 - val_loss: 0.1023 - val_acc: 0.9791\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1241c8a20>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data= (X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1132e1978>]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1132e1b38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"'triangular2' Policy Reset at 20000 Iterations\")\n",
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular._reset()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.1205 - acc: 0.9735 - val_loss: 0.2254 - val_acc: 0.9611\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.2474 - acc: 0.9510 - val_loss: 0.1490 - val_acc: 0.9699\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.1248 - acc: 0.9708 - val_loss: 0.1467 - val_acc: 0.9715\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.1138 - acc: 0.9739 - val_loss: 0.1150 - val_acc: 0.9762\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0825 - acc: 0.9793 - val_loss: 0.1136 - val_acc: 0.9766\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0766 - acc: 0.9804 - val_loss: 0.1132 - val_acc: 0.9759\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0711 - acc: 0.9824 - val_loss: 0.1118 - val_acc: 0.9782\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 10s 161us/step - loss: 0.0653 - acc: 0.9827 - val_loss: 0.1142 - val_acc: 0.9779\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0602 - acc: 0.9841 - val_loss: 0.1148 - val_acc: 0.9774\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0612 - acc: 0.9836 - val_loss: 0.1161 - val_acc: 0.9784\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1241c20b8>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data=(X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x113187c88>]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1131870f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"'triangular2' Policy Reset at 20000 Iterations\")\n",
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular = CyclicLR(mode='exp_range', step_size=np.ceil(X_train.shape[0]/batch_size), gamma=0.9994)\n",
    "\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 10s 168us/step - loss: 0.0912 - acc: 0.9781 - val_loss: 0.1623 - val_acc: 0.9706\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.1049 - acc: 0.9763 - val_loss: 0.1222 - val_acc: 0.9771\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0735 - acc: 0.9816 - val_loss: 0.1361 - val_acc: 0.9750\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0758 - acc: 0.9814 - val_loss: 0.1103 - val_acc: 0.9778\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0602 - acc: 0.9847 - val_loss: 0.1192 - val_acc: 0.9786\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0587 - acc: 0.9845 - val_loss: 0.1102 - val_acc: 0.9797\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0496 - acc: 0.9872 - val_loss: 0.1100 - val_acc: 0.9801\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0496 - acc: 0.9876 - val_loss: 0.1112 - val_acc: 0.9788\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.0469 - acc: 0.9879 - val_loss: 0.1082 - val_acc: 0.9800\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 10s 163us/step - loss: 0.0436 - acc: 0.9884 - val_loss: 0.1116 - val_acc: 0.9794\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1241c2978>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data=(X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x12f9c95f8>]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12f9c96a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"CLR - 'exp_range' Policy\")\n",
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular._reset(new_base_lr=0.003, new_step_size=np.ceil(X_train.shape[0]/batch_size), new_max_lr=0.009)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 10s 161us/step - loss: 0.1132 - acc: 0.9752 - val_loss: 0.1731 - val_acc: 0.9685\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 10s 161us/step - loss: 0.1447 - acc: 0.9690 - val_loss: 0.1257 - val_acc: 0.9774\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 10s 162us/step - loss: 0.0855 - acc: 0.9787 - val_loss: 0.1174 - val_acc: 0.9778\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.0825 - acc: 0.9800 - val_loss: 0.1154 - val_acc: 0.9775\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.0708 - acc: 0.9824 - val_loss: 0.1243 - val_acc: 0.9769\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 10s 158us/step - loss: 0.0763 - acc: 0.9817 - val_loss: 0.1128 - val_acc: 0.9783\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 10s 161us/step - loss: 0.0627 - acc: 0.9837 - val_loss: 0.1222 - val_acc: 0.9771\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 10s 158us/step - loss: 0.0582 - acc: 0.9851 - val_loss: 0.1092 - val_acc: 0.9797\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.0573 - acc: 0.9854 - val_loss: 0.1379 - val_acc: 0.9781\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 9s 156us/step - loss: 0.0547 - acc: 0.9858 - val_loss: 0.1201 - val_acc: 0.9788\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1132a1160>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data=(X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x12fff5400>]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12549cef0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr = lambda x: 0.5*(1+np.sin(x*np.pi/2.))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular = CyclicLR(scale_fn=clr, scale_mode='cycle', step_size=np.ceil(X_train.shape[0]/batch_size))\n",
    "\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.0793 - acc: 0.9824 - val_loss: 0.1891 - val_acc: 0.9683\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1474 - acc: 0.9707 - val_loss: 0.1252 - val_acc: 0.9764\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0828 - acc: 0.9814 - val_loss: 0.1637 - val_acc: 0.9760\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 9s 156us/step - loss: 0.0777 - acc: 0.9827 - val_loss: 0.1341 - val_acc: 0.9800\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0604 - acc: 0.9864 - val_loss: 0.1339 - val_acc: 0.9801\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0576 - acc: 0.9864 - val_loss: 0.1337 - val_acc: 0.9800\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 9s 156us/step - loss: 0.0610 - acc: 0.9856 - val_loss: 0.1266 - val_acc: 0.9783\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0714 - acc: 0.9833 - val_loss: 0.1209 - val_acc: 0.9792\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 10s 161us/step - loss: 0.0839 - acc: 0.9810 - val_loss: 0.2484 - val_acc: 0.9665\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.2096 - acc: 0.9588 - val_loss: 0.1443 - val_acc: 0.9734\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x12410f2b0>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data=(X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x130f707b8>]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x130f70780>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"CLR - Custom Sinusoidal Cycle-Policy\")\n",
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr = lambda x: 1/(5**(x*0.0001))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "clr_triangular = CyclicLR(scale_fn=clr, scale_mode='iterations', step_size=np.ceil(X_train.shape[0]/batch_size))\n",
    "\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 10s 159us/step - loss: 0.1319 - acc: 0.9733 - val_loss: 0.2038 - val_acc: 0.9695\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.1326 - acc: 0.9731 - val_loss: 0.1458 - val_acc: 0.9751\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0902 - acc: 0.9800 - val_loss: 0.1846 - val_acc: 0.9709\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.1127 - acc: 0.9772 - val_loss: 0.1334 - val_acc: 0.9783\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0860 - acc: 0.9794 - val_loss: 0.1398 - val_acc: 0.9784\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 10s 160us/step - loss: 0.0905 - acc: 0.9804 - val_loss: 0.1285 - val_acc: 0.9785\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0681 - acc: 0.9836 - val_loss: 0.1533 - val_acc: 0.9743\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 9s 157us/step - loss: 0.0735 - acc: 0.9836 - val_loss: 0.1421 - val_acc: 0.9798\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 9s 158us/step - loss: 0.0603 - acc: 0.9864 - val_loss: 0.1352 - val_acc: 0.9776\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 10s 161us/step - loss: 0.0649 - acc: 0.9856 - val_loss: 0.1335 - val_acc: 0.9798\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x12ffe8a90>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, batch_size=128, epochs=10, validation_data=(X_test, y_test), callbacks=[clr_triangular])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x13137c2e8>]"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x13137c2b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.xlabel('Training Iterations')\n",
    "plt.ylabel('Learning Rate')\n",
    "plt.title(\"CLR - Custom Iteration-Policy\")\n",
    "plt.plot(clr_triangular.history['iterations'], clr_triangular.history['lr'])"
   ]
  },
  {
   "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": "3.5.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}