cae_mnist.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = [1, 2, 3]\n",
    "a[-3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from utils import *\n",
    "from modules import *\n",
    "from models import *\n",
    "from torchvision import datasets, transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.cuda.empty_cache()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_data = datasets.MNIST(root='/Users/Linsu Han/Documents/Data/mnist/train.pt', train=True, download=True, transform=transforms.ToTensor())\n",
    "val_data = datasets.MNIST(root='/Users/Linsu Han/Documents/Data/mnist/val.pt', train=False, download=True, transform=transforms.ToTensor())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "in_distribution = {0, 1, 2, 3, 4}\n",
    "train_subset = get_subset(train_data, in_distribution)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_data 60000\n",
      "train_subset 30596\n",
      "val_data 10000\n"
     ]
    }
   ],
   "source": [
    "print('train_data', len(train_data))\n",
    "print('train_subset', len(train_subset))\n",
    "print('val_data', len(val_data))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Conv2d\n",
    "\n",
    "`nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros')`\n",
    "\n",
    "$$\n",
    "\\text{out}(N_i, C_{\\text{out}_j}) = \\text{bias}(C_{\\text{out}_j}) + \\sum_{k = 0}^{C_{\\text{in}} - 1} \\text{weight}(C_{\\text{out}_j}, k) \\star \\text{input}(N_i, k)\n",
    "$$\n",
    "\n",
    "$$\n",
    "H_{out} = \\left\\lfloor\\frac{H_{in}  + 2 \\times \\text{padding}[0] - \\text{dilation}[0]\n",
    "                        \\times (\\text{kernel\\_size}[0] - 1) - 1}{\\text{stride}[0]} + 1\\right\\rfloor\n",
    "$$\n",
    "\n",
    "$$\n",
    "W_{out} = \\left\\lfloor\\frac{W_{in}  + 2 \\times \\text{padding}[1] - \\text{dilation}[1]\n",
    "        \\times (\\text{kernel\\_size}[1] - 1) - 1}{\\text{stride}[1]} + 1\\right\\rfloor\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MaxPool2d\n",
    "\n",
    "`nn.MaxPool2d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False)`\n",
    "\n",
    "$$\n",
    "H_{out} = \\left\\lfloor\\frac{H_{in} + 2 * \\text{padding[0]} - \\text{dilation[0]} \\times (\\text{kernel\\_size[0]} - 1) - 1}{\\text{stride[0]}} + 1\\right\\rfloor\n",
    "$$\n",
    "\n",
    "$$\n",
    "W_{out} = \\left\\lfloor\\frac{W_{in} + 2 * \\text{padding[1]} - \\text{dilation[1]} \\times (\\text{kernel\\_size[1]} - 1) - 1}{\\text{stride[1]}} + 1\\right\\rfloor\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# ConvTranspose2d\n",
    "`nn.ConvTranspose2d(in_channels, out_channels, kernel_size, stride=1, padding=0, output_padding=0, groups=1, bias=True, dilation=1, padding_mode='zeros')`\n",
    "\n",
    "$$\n",
    "              H_{out} = (H_{in} - 1) \\times \\text{stride}[0] - 2 \\times \\text{padding}[0] + \\text{dilation}[0]\n",
    "                        \\times (\\text{kernel\\_size}[0] - 1) + \\text{output\\_padding}[0] + 1\n",
    "$$\n",
    "\n",
    "$$\n",
    "              W_{out} = (W_{in} - 1) \\times \\text{stride}[1] - 2 \\times \\text{padding}[1] + \\text{dilation}[1]\n",
    "                        \\times (\\text{kernel\\_size}[1] - 1) + \\text{output\\_padding}[1] + 1\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Defaults\n",
    "# nn.Sequential(*args)\n",
    "# nn.Conv2d(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros')\n",
    "# nn.MaxPool2d(kernel_size, stride=None, padding=0, dilation=1, return_indices=False, ceil_mode=False)\n",
    "# nn.Linear(in_features, out_features, bias=True)\n",
    "# nn.ConvTranspose2d(in_channels, out_channels, kernel_size, stride=1, padding=0, output_padding=0, groups=1, bias=True, dilation=1, padding_mode='zeros')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Output: 24\n",
      "Output: 22\n",
      "Output: 20\n",
      "Output: 22\n",
      "Output: 24\n",
      "Output: 28\n"
     ]
    }
   ],
   "source": [
    "# Using stride=1\n",
    "dim = 28\n",
    "dim = conv_dim(dim, 5, 1, 0, 1)\n",
    "dim = conv_dim(dim, 3, 1, 0, 1)\n",
    "dim = conv_dim(dim, 3, 1, 0, 1)\n",
    "dim = deconv_dim(dim, 3, 1, 0, 0, 1)\n",
    "dim = deconv_dim(dim, 3, 1, 0, 0, 1)\n",
    "dim = deconv_dim(dim, 5, 1, 0, 0, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`f(x) = floor[ (x + 2*padding1 - kernel1) / stride1 + 1]`\n",
    "\n",
    "`g(x) = (x - 1)*stride2 - 2*padding2 + kernel2`\n",
    "\n",
    "Given `stride1, stride2 > 1`, is it possible to have parameters s.t. `g(f(x)) = x` for all `x`?\n",
    "\n",
    "=> Yes iff `x + 2*padding1 - kernel1` is divisible by `stride1`\n",
    "\n",
    "Try:\n",
    "\n",
    "let `padding1 = 0`\n",
    "\n",
    "need to find `kernel1` s.t. `(x-kernel1) % stride1 == 0` for all `x`\n",
    "\n",
    "We can control `kernel1` s.t. numerator is even so it works for `stride1 = 2`\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Output: 13\n",
      "Output: 6\n",
      "Output: 2\n",
      "Output: 6\n",
      "Output: 13\n",
      "Output: 28\n"
     ]
    }
   ],
   "source": [
    "# Using stride=2\n",
    "\n",
    "dim = 28\n",
    "\n",
    "def get_kernel(dim):\n",
    "    if dim%2==0: # even input\n",
    "        return 4 # subtract by even\n",
    "    else: # odd input\n",
    "        return 3 # subtract by odd\n",
    "\n",
    "k1 = get_kernel(dim)\n",
    "dim = conv_dim(dim, k1, 2, 0, 1)\n",
    "k2 = get_kernel(dim)\n",
    "dim = conv_dim(dim, k2, 2, 0, 1)\n",
    "k3 = get_kernel(dim)\n",
    "dim = conv_dim(dim, k3, 2, 0, 1)\n",
    "dim = deconv_dim(dim, k3, 2, 0, 0, 1)\n",
    "dim = deconv_dim(dim, k2, 2, 0, 0, 1)\n",
    "dim = deconv_dim(dim, k1, 2, 0, 0, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "nn_module = ConvAutoencoder(16, *train_data[0][0].shape, hid=[4, 8, 16], activation=nn.ReLU())\n",
    "model = CAEModel(nn_module)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  6%|█████▏                                                                             | 1/16 [00:09<02:16,  9.12s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.03944852509095004\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 12%|██████████▍                                                                        | 2/16 [00:15<01:48,  7.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.01817514025387252\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 19%|███████████████▌                                                                   | 3/16 [00:22<01:34,  7.29s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.015579405301051018\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 25%|████████████████████▊                                                              | 4/16 [00:29<01:24,  7.05s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.013866190809285504\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 31%|█████████████████████████▉                                                         | 5/16 [00:36<01:16,  6.94s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.012417072420521341\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 38%|███████████████████████████████▏                                                   | 6/16 [00:42<01:08,  6.86s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.01096379877843718\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 44%|████████████████████████████████████▎                                              | 7/16 [00:49<01:01,  6.81s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.009825521666467066\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 50%|█████████████████████████████████████████▌                                         | 8/16 [00:56<00:54,  6.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.009315237360096526\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 56%|██████████████████████████████████████████████▋                                    | 9/16 [01:03<00:47,  6.80s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.008929697765560983\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 62%|███████████████████████████████████████████████████▎                              | 10/16 [01:09<00:40,  6.77s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.008612219202395455\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 69%|████████████████████████████████████████████████████████▍                         | 11/16 [01:16<00:33,  6.75s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.008324891720998685\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 75%|█████████████████████████████████████████████████████████████▌                    | 12/16 [01:23<00:27,  6.75s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.008114807655628043\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 81%|██████████████████████████████████████████████████████████████████▋               | 13/16 [01:30<00:20,  6.75s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.007910612476220824\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|███████████████████████████████████████████████████████████████████████▊          | 14/16 [01:36<00:13,  6.74s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.007716913432634144\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 94%|████████████████████████████████████████████████████████████████████████████▉     | 15/16 [01:43<00:06,  6.72s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.007599543629247255\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 16/16 [01:50<00:00,  6.88s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.007445675336980614\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "model.train(train_subset, 16, verbose=True, batch_size=32, num_workers=2, shuffle=True)\n",
    "model.save('./mnist_cae.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# model.load('./mnist_cae.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x19b11ab0a30>]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(model.losses_epoch)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "outputs = model.evaluate([val_data])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "mse_dict = {k:[d['mse'] for d in v] for k, v in outputs.items()}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x19b57b94430>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(mse_dict.keys(), [sum(v)/len(v) for v in mse_dict.values()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1800x288 with 16 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model.visualize(val_data, 8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "thresholds = np.arange(1e-3, 50e-3, 1e-3)\n",
    "tpr, tnr = get_acc(mse_dict, thresholds, in_distribution)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x19ba955f1f0>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(thresholds, tpr)\n",
    "plt.plot(thresholds, tnr)\n",
    "plt.legend(('tpr', 'tnr'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'True Positive Rate')"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(1-np.array(tnr), tpr)\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.title('ROC Curve')\n",
    "plt.xlabel('False Postive Rate')\n",
    "plt.ylabel('True Positive Rate')"
   ]
  }
 ],
 "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": "3.8.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}