matrix-factorization.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras import layers\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.regularizers import l2\n",
    "from tensorflow.keras.optimizers import SGD, Adam\n",
    "from tensorflow.keras.metrics import MeanSquaredError, RootMeanSquaredError"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "outputs": [],
   "source": [
    "from sklearn.utils import shuffle"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "outputs": [],
   "source": [
    "df = pd.read_csv('data/ratings_preprocessed.csv', index_col=0)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "outputs": [
    {
     "data": {
      "text/plain": "   rating  userId  movieId\n0     4.0     193     1744\n1     4.5     417     1744\n2     3.5     914     1744\n3     3.0     813      572\n4     2.5     193      261",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>rating</th>\n      <th>userId</th>\n      <th>movieId</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>4.0</td>\n      <td>193</td>\n      <td>1744</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>4.5</td>\n      <td>417</td>\n      <td>1744</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>3.5</td>\n      <td>914</td>\n      <td>1744</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>3.0</td>\n      <td>813</td>\n      <td>572</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>2.5</td>\n      <td>193</td>\n      <td>261</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "outputs": [],
   "source": [
    "df = df[(df['movieId'] < 2000) & (df['userId'] < 8000)]"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of ratings in the dataset is 5392234\n",
      "The number of users in the dataset is 10000\n",
      "The number of movies in the dataset is 2000\n"
     ]
    }
   ],
   "source": [
    "print(f'The number of ratings in the dataset is {df.shape[0]}')\n",
    "print(f'The number of users in the dataset is {df.userId.nunique()}')\n",
    "print(f'The number of movies in the dataset is {df.movieId.nunique()}')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "outputs": [],
   "source": [
    "N = df.userId.max() + 1\n",
    "M = df.movieId.max() + 1"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "outputs": [],
   "source": [
    "df = shuffle(df)\n",
    "cutoff = int(0.8*len(df))\n",
    "df_train = df.iloc[:cutoff]\n",
    "df_test = df.iloc[cutoff:]"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of users in the training dataset is 4313787\n",
      "The number of users in the test dataset is 1078447\n"
     ]
    }
   ],
   "source": [
    "print(f'The number of users in the training dataset is {df_train.shape[0]}')\n",
    "print(f'The number of users in the test dataset is {df_test.shape[0]}')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "outputs": [],
   "source": [
    "K = 20\n",
    "mu = df_train.rating.mean()\n",
    "epochs = 25\n",
    "reg = 0. # regularization penalty"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "markdown",
   "source": [
    "**Building Layers of the Neural Network**"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metal device set to: Apple M1 Pro\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-25 21:30:01.190323: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:306] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.\n",
      "2022-12-25 21:30:01.190528: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:272] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: <undefined>)\n"
     ]
    }
   ],
   "source": [
    "u = layers.Input(shape=(1, ))\n",
    "m = layers.Input(shape=(1, ))\n",
    "u_embedding = layers.Embedding(N, K, embeddings_regularizer=l2(reg))(u)\n",
    "m_embedding = layers.Embedding(M, K, embeddings_regularizer=l2(reg))(m)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "outputs": [],
   "source": [
    "u_bias = layers.Embedding(N, 1, embeddings_regularizer=l2(reg))(u)\n",
    "m_bias = layers.Embedding(M, 1, embeddings_regularizer=l2(reg))(m)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "outputs": [],
   "source": [
    "x = layers.Dot(axes=2)([u_embedding, m_embedding])"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "outputs": [],
   "source": [
    "x = layers.Add()([x, u_bias, m_bias])"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "outputs": [],
   "source": [
    "x = layers.Flatten()(x)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "outputs": [],
   "source": [
    "model = Model(inputs=[u, m], outputs=x)"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_1 (InputLayer)           [(None, 1)]          0           []                               \n",
      "                                                                                                  \n",
      " input_2 (InputLayer)           [(None, 1)]          0           []                               \n",
      "                                                                                                  \n",
      " embedding (Embedding)          (None, 1, 20)        200000      ['input_1[0][0]']                \n",
      "                                                                                                  \n",
      " embedding_1 (Embedding)        (None, 1, 20)        40000       ['input_2[0][0]']                \n",
      "                                                                                                  \n",
      " dot (Dot)                      (None, 1, 1)         0           ['embedding[0][0]',              \n",
      "                                                                  'embedding_1[0][0]']            \n",
      "                                                                                                  \n",
      " embedding_2 (Embedding)        (None, 1, 1)         10000       ['input_1[0][0]']                \n",
      "                                                                                                  \n",
      " embedding_3 (Embedding)        (None, 1, 1)         2000        ['input_2[0][0]']                \n",
      "                                                                                                  \n",
      " add (Add)                      (None, 1, 1)         0           ['dot[0][0]',                    \n",
      "                                                                  'embedding_2[0][0]',            \n",
      "                                                                  'embedding_3[0][0]']            \n",
      "                                                                                                  \n",
      " flatten (Flatten)              (None, 1)            0           ['add[0][0]']                    \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 252,000\n",
      "Trainable params: 252,000\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "outputs": [],
   "source": [
    "model.compile(\n",
    "    loss='mse',\n",
    "    optimizer=SGD(learning_rate=0.01, momentum=0.9),\n",
    "    metrics=[MeanSquaredError()]\n",
    ")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/25\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-25 21:30:16.716488: W tensorflow/core/platform/profile_utils/cpu_utils.cc:128] Failed to get CPU frequency: 0 Hz\n",
      "2022-12-25 21:30:16.984230: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26962/26962 [==============================] - ETA: 0s - loss: 0.8000 - mean_squared_error: 0.8000"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-25 21:33:19.656101: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26962/26962 [==============================] - 206s 8ms/step - loss: 0.8000 - mean_squared_error: 0.8000 - val_loss: 0.7207 - val_mean_squared_error: 0.7207\n",
      "Epoch 2/25\n",
      "26962/26962 [==============================] - 204s 8ms/step - loss: 0.7016 - mean_squared_error: 0.7016 - val_loss: 0.6870 - val_mean_squared_error: 0.6870\n",
      "Epoch 3/25\n",
      "26962/26962 [==============================] - 200s 7ms/step - loss: 0.6811 - mean_squared_error: 0.6811 - val_loss: 0.6761 - val_mean_squared_error: 0.6761\n",
      "Epoch 4/25\n",
      "26962/26962 [==============================] - 321s 12ms/step - loss: 0.6737 - mean_squared_error: 0.6737 - val_loss: 0.6720 - val_mean_squared_error: 0.6720\n",
      "Epoch 5/25\n",
      "26962/26962 [==============================] - 202s 7ms/step - loss: 0.6706 - mean_squared_error: 0.6706 - val_loss: 0.6703 - val_mean_squared_error: 0.6703\n",
      "Epoch 6/25\n",
      "26962/26962 [==============================] - 201s 7ms/step - loss: 0.6693 - mean_squared_error: 0.6693 - val_loss: 0.6695 - val_mean_squared_error: 0.6695\n",
      "Epoch 7/25\n",
      "26962/26962 [==============================] - 210s 8ms/step - loss: 0.6685 - mean_squared_error: 0.6685 - val_loss: 0.6692 - val_mean_squared_error: 0.6692\n",
      "Epoch 8/25\n",
      "26962/26962 [==============================] - 207s 8ms/step - loss: 0.6681 - mean_squared_error: 0.6681 - val_loss: 0.6689 - val_mean_squared_error: 0.6689\n",
      "Epoch 9/25\n",
      "26962/26962 [==============================] - 208s 8ms/step - loss: 0.6677 - mean_squared_error: 0.6677 - val_loss: 0.6687 - val_mean_squared_error: 0.6687\n",
      "Epoch 10/25\n",
      "26962/26962 [==============================] - 200s 7ms/step - loss: 0.6672 - mean_squared_error: 0.6672 - val_loss: 0.6683 - val_mean_squared_error: 0.6683\n",
      "Epoch 11/25\n",
      "26962/26962 [==============================] - 201s 7ms/step - loss: 0.6666 - mean_squared_error: 0.6666 - val_loss: 0.6677 - val_mean_squared_error: 0.6677\n",
      "Epoch 12/25\n",
      "26962/26962 [==============================] - 202s 7ms/step - loss: 0.6655 - mean_squared_error: 0.6655 - val_loss: 0.6665 - val_mean_squared_error: 0.6665\n",
      "Epoch 13/25\n",
      "26962/26962 [==============================] - 207s 8ms/step - loss: 0.6638 - mean_squared_error: 0.6638 - val_loss: 0.6646 - val_mean_squared_error: 0.6646\n",
      "Epoch 14/25\n",
      "26962/26962 [==============================] - 206s 8ms/step - loss: 0.6609 - mean_squared_error: 0.6609 - val_loss: 0.6615 - val_mean_squared_error: 0.6615\n",
      "Epoch 15/25\n",
      "26962/26962 [==============================] - 203s 8ms/step - loss: 0.6565 - mean_squared_error: 0.6565 - val_loss: 0.6567 - val_mean_squared_error: 0.6567\n",
      "Epoch 16/25\n",
      "26962/26962 [==============================] - 183s 7ms/step - loss: 0.6502 - mean_squared_error: 0.6502 - val_loss: 0.6503 - val_mean_squared_error: 0.6503\n",
      "Epoch 17/25\n",
      "26962/26962 [==============================] - 204s 8ms/step - loss: 0.6426 - mean_squared_error: 0.6426 - val_loss: 0.6430 - val_mean_squared_error: 0.6430\n",
      "Epoch 18/25\n",
      "26962/26962 [==============================] - 205s 8ms/step - loss: 0.6345 - mean_squared_error: 0.6345 - val_loss: 0.6361 - val_mean_squared_error: 0.6361\n",
      "Epoch 19/25\n",
      "26962/26962 [==============================] - 204s 8ms/step - loss: 0.6270 - mean_squared_error: 0.6270 - val_loss: 0.6299 - val_mean_squared_error: 0.6299\n",
      "Epoch 20/25\n",
      "26962/26962 [==============================] - 203s 8ms/step - loss: 0.6202 - mean_squared_error: 0.6202 - val_loss: 0.6245 - val_mean_squared_error: 0.6245\n",
      "Epoch 21/25\n",
      "26962/26962 [==============================] - 206s 8ms/step - loss: 0.6138 - mean_squared_error: 0.6138 - val_loss: 0.6191 - val_mean_squared_error: 0.6191\n",
      "Epoch 22/25\n",
      "26962/26962 [==============================] - 204s 8ms/step - loss: 0.6074 - mean_squared_error: 0.6074 - val_loss: 0.6139 - val_mean_squared_error: 0.6139\n",
      "Epoch 23/25\n",
      "26962/26962 [==============================] - 209s 8ms/step - loss: 0.6008 - mean_squared_error: 0.6008 - val_loss: 0.6086 - val_mean_squared_error: 0.6086\n",
      "Epoch 24/25\n",
      "26962/26962 [==============================] - 210s 8ms/step - loss: 0.5940 - mean_squared_error: 0.5940 - val_loss: 0.6034 - val_mean_squared_error: 0.6034\n",
      "Epoch 25/25\n",
      "26962/26962 [==============================] - 211s 8ms/step - loss: 0.5873 - mean_squared_error: 0.5873 - val_loss: 0.5983 - val_mean_squared_error: 0.5983\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "    x = [df_train['userId'].values, df_train['movieId'].values],\n",
    "    y = df_train['rating'].values - mu,\n",
    "    epochs = epochs,\n",
    "    batch_size = 128,\n",
    "    validation_split=0.2\n",
    ")"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'], label=\"train loss\")\n",
    "plt.plot(history.history['val_loss'], label=\"validation loss\")\n",
    "plt.legend()\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 640x480 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['mean_squared_error'], label=\"train mse\")\n",
    "plt.plot(history.history['val_mean_squared_error'], label=\"validation mse\")\n",
    "plt.legend()\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   49/33702 [..............................] - ETA: 1:48"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-12-25 22:57:45.067993: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:114] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "33702/33702 [==============================] - 80s 2ms/step\n",
      "The Trianing MSE is 0.584852933883667\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(\n",
    "    [df_train.userId.values, df_train.movieId.values],\n",
    "    batch_size=128\n",
    ")\n",
    "\n",
    "y_true = df_train.rating.values - mu\n",
    "mse = MeanSquaredError()\n",
    "print(f'The Trianing MSE is {mse(y_true, y_pred).numpy()}')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "33702/33702 [==============================] - 82s 2ms/step\n",
      "The test MSE is 0.6023701429367065\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(\n",
    "    [df_test.userId.values, df_test.movieId.values]\n",
    ")\n",
    "\n",
    "y_true = df_test.rating.values - mu\n",
    "mse = MeanSquaredError()\n",
    "print(f'The test MSE is {mse(y_true, y_pred).numpy()}')"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "outputs": [],
   "source": [
    "from tensorflow.keras import backend as K"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "outputs": [],
   "source": [
    "K.clear_session()"
   ],
   "metadata": {
    "collapsed": false
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}