docs/vignettes/Stochastic Wasserstein Singular Vectors.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "# Stochastic Wasserstein Singular Vectors"
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "This Jupyter Notebook will walk you through an easy example of stochastic computation of Wasserstein Singular Vectors. This example is small enough to be run on CPU."
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Imports"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "source": [
    "import wsingular\n",
    "import torch\n",
    "import matplotlib.pyplot as plt"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stderr",
     "text": [
      "<frozen importlib._bootstrap>:219: RuntimeWarning: scipy._lib.messagestream.MessageStream size changed, may indicate binary incompatibility. Expected 56 from C header, got 64 from PyObject\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Generate toy data"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "source": [
    "# Define the dtype and device to work with.\n",
    "dtype = torch.double\n",
    "device = \"cpu\""
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "source": [
    "# Define the dimensions of our problem.\n",
    "n_samples = 20\n",
    "n_features = 30"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "source": [
    "# Initialize an empty dataset.\n",
    "dataset = torch.zeros((n_samples, n_features), dtype=dtype)\n",
    "\n",
    "# Iterate over the features and samples.\n",
    "for i in range(n_samples):\n",
    "    for j in range(n_features):\n",
    "\n",
    "        # Fill the dataset with translated histograms.\n",
    "        dataset[i, j] = i/n_samples - j/n_features\n",
    "        dataset[i, j] = torch.abs(dataset[i, j] % 1)\n",
    "\n",
    "# Take the distance to 0 on the torus.\n",
    "dataset = torch.min(dataset, 1 - dataset)\n",
    "\n",
    "# Make it a guassian.\n",
    "dataset = torch.exp(-(dataset**2) / 0.1)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "source": [
    "# Plot the dataset.\n",
    "plt.title('The dataset')\n",
    "plt.imshow(dataset)\n",
    "plt.colorbar()\n",
    "plt.xticks([])\n",
    "plt.yticks([])\n",
    "plt.show()"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Compute the WSV"
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "source": [
    "# Compute the WSV.\n",
    "C, D = wsingular.stochastic_wasserstein_singular_vectors(\n",
    "    dataset,\n",
    "    dtype=dtype,\n",
    "    device=device,\n",
    "    n_iter=1_000,\n",
    "    sample_prop=1e-1,\n",
    ")\n"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "source": [
    "# Display the WSV.\n",
    "fig, axes = plt.subplots(1, 2, figsize=(10, 5))\n",
    "fig.suptitle('Wasserstein Singular Vectors')\n",
    "\n",
    "axes[0].set_title('Distance between samples.')\n",
    "axes[0].imshow(D)\n",
    "axes[0].set_xticks(range(0, n_samples, 5))\n",
    "axes[0].set_yticks(range(0, n_samples, 5))\n",
    "\n",
    "axes[1].set_title('Distance between features.')\n",
    "axes[1].imshow(C)\n",
    "axes[1].set_xticks(range(0, n_features, 5))\n",
    "axes[1].set_yticks(range(0, n_features, 5))\n",
    "\n",
    "plt.show()"
   ],
   "outputs": [
    {
     "output_type": "display_data",
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "source": [
    "A, B = wsingular.utils.normalize_dataset(dataset, dtype=dtype, device=device)\n",
    "wsingular.utils.check_uniqueness(A, B, C, D)"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "metadata": {},
     "execution_count": 8
    }
   ],
   "metadata": {}
  }
 ],
 "metadata": {
  "orig_nbformat": 4,
  "language_info": {
   "name": "python",
   "version": "3.8.5",
   "mimetype": "text/x-python",
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "pygments_lexer": "ipython3",
   "nbconvert_exporter": "python",
   "file_extension": ".py"
  },
  "kernelspec": {
   "name": "python3",
   "display_name": "Python 3.8.5 64-bit ('base': conda)"
  },
  "interpreter": {
   "hash": "0bbb79805e160b9a386126d782961d35b74be8ad2fd67e519c2698cde08d84ae"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}