{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Importing the libraries\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.parallel\n",
    "import torch.optim as optim\n",
    "import torch.utils.data\n",
    "from torch.autograd import Variable"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Importing the dataset\n",
    "movies = pd.read_csv('ml-1m/movies.dat', sep = '::', header = None, engine = 'python', encoding = 'latin-1')\n",
    "users = pd.read_csv('ml-1m/users.dat', sep = '::', header = None, engine = 'python', encoding = 'latin-1')\n",
    "ratings = pd.read_csv('ml-1m/ratings.dat', sep = '::', header = None, engine = 'python', encoding = 'latin-1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preparing the training set and the test set\n",
    "training_set = pd.read_csv('ml-100k/u1.base', delimiter = '\\t')\n",
    "training_set = np.array(training_set, dtype = 'int')\n",
    "test_set = pd.read_csv('ml-100k/u1.test', delimiter = '\\t')\n",
    "test_set = np.array(test_set, dtype = 'int')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Getting the number of users and movies\n",
    "nb_users = int(max(max(training_set[:,0]), max(test_set[:,0])))\n",
    "nb_movies = int(max(max(training_set[:,1]), max(test_set[:,1])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Converting the data into an array with users in lines and movies in columns\n",
    "def convert(data):\n",
    "    new_data = []\n",
    "    for id_users in range(1, nb_users + 1):\n",
    "        id_movies = data[:,1][data[:,0] == id_users]\n",
    "        id_ratings = data[:,2][data[:,0] == id_users]\n",
    "        ratings = np.zeros(nb_movies)\n",
    "        ratings[id_movies - 1] = id_ratings\n",
    "        new_data.append(list(ratings))\n",
    "    return new_data\n",
    "training_set = convert(training_set)\n",
    "test_set = convert(test_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Converting the data into Torch tensors\n",
    "training_set = torch.FloatTensor(training_set)\n",
    "test_set = torch.FloatTensor(test_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Creating the architecture of the Neural Network\n",
    "class SAE(nn.Module):\n",
    "    def __init__(self, ):\n",
    "        super(SAE, self).__init__()\n",
    "        self.fc1 = nn.Linear(nb_movies, 20)\n",
    "        self.fc2 = nn.Linear(20, 10)\n",
    "        self.fc3 = nn.Linear(10, 20)\n",
    "        self.fc4 = nn.Linear(20, nb_movies)\n",
    "        self.activation = nn.Sigmoid()\n",
    "    def forward(self, x):\n",
    "        x = self.activation(self.fc1(x))\n",
    "        x = self.activation(self.fc2(x))\n",
    "        x = self.activation(self.fc3(x))\n",
    "        x = self.fc4(x)\n",
    "        return x\n",
    "sae = SAE()\n",
    "criterion = nn.MSELoss()\n",
    "optimizer = optim.RMSprop(sae.parameters(), lr = 0.01, weight_decay = 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Training the SAE\n",
    "nb_epoch = 200\n",
    "for epoch in range(1, nb_epoch + 1):\n",
    "    train_loss = 0\n",
    "    s = 0.\n",
    "    for id_user in range(nb_users):\n",
    "        input = Variable(training_set[id_user]).unsqueeze(0)\n",
    "        target = input.clone()\n",
    "        if torch.sum(target.data > 0) > 0:\n",
    "            output = sae(input)\n",
    "            target.require_grad = False\n",
    "            output[target == 0] = 0\n",
    "            loss = criterion(output, target)\n",
    "            mean_corrector = nb_movies/float(torch.sum(target.data > 0) + 1e-10)\n",
    "            loss.backward()\n",
    "            train_loss += np.sqrt(loss.item()*mean_corrector)\n",
    "            s += 1.\n",
    "            optimizer.step()\n",
    "    print('epoch: '+str(epoch)+' loss: '+str(train_loss/s))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Testing the SAE\n",
    "test_loss = 0\n",
    "s = 0.\n",
    "for id_user in range(nb_users):\n",
    "    input = Variable(training_set[id_user]).unsqueeze(0)\n",
    "    target = Variable(test_set[id_user])\n",
    "    if torch.sum(target.data > 0) > 0:\n",
    "        output = sae(input)\n",
    "        target.require_grad = False\n",
    "        output[0, target == 0] = 0\n",
    "        loss = criterion(output, target)\n",
    "        mean_corrector = nb_movies/float(torch.sum(target.data > 0) + 1e-10)\n",
    "        test_loss += np.sqrt(loss.item()*mean_corrector)\n",
    "        s += 1.\n",
    "print('test loss: '+str(test_loss/s))"
   ]
  }
 ],
 "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": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}