docs/nbs/BNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "85f6ecbd",
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import jax.numpy as jnp \n",
    "from jax import random, jit\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2553b708",
   "metadata": {},
   "source": [
    "# Bayesian Neural Network\n",
    "\n",
    "- Use an SGLD kernel and compute the accuracy on the test dataset while training\n",
    "- MNIST dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "fa670e8b",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:absl:No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)\n"
     ]
    }
   ],
   "source": [
    "from sgmcmcjax.models.bayesian_NN.NN_data import X_train, X_test, y_train, y_test\n",
    "from sgmcmcjax.models.bayesian_NN.NN_model import init_network, loglikelihood, logprior, accuracy\n",
    "\n",
    "from sgmcmcjax.kernels import build_sgld_kernel\n",
    "from tqdm.auto import tqdm\n",
    "\n",
    "batch_size = int(0.01*X_train.shape[0])\n",
    "data = (X_train, y_train)\n",
    "\n",
    "init_fn, my_kernel, get_params = build_sgld_kernel(5e-5, loglikelihood, logprior, data, batch_size)\n",
    "\n",
    "my_kernel = jit(my_kernel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "cd0a9011",
   "metadata": {},
   "outputs": [],
   "source": [
    "# define the inital state\n",
    "key = random.PRNGKey(10)\n",
    "key, subkey1, subkey2 = random.split(key,3)\n",
    "sizes = [784, 100, 10]\n",
    "params_IC = init_network(subkey1, sizes)\n",
    "state = init_fn(subkey2, params_IC)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "afbb38dd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8ebfd5f5e6854582b46af9ce43684f2c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 29.5 s, sys: 2.79 s, total: 32.3 s\n",
      "Wall time: 12.9 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "Nsamples = 2000\n",
    "samples = []\n",
    "accuracy_list = []\n",
    "\n",
    "for i in tqdm(range(Nsamples)):\n",
    "    key, subkey = random.split(key)\n",
    "    state = my_kernel(i, subkey, state)\n",
    "    samples.append(get_params(state))\n",
    "    if i%10==0:\n",
    "        accuracy_list.append(accuracy(get_params(state), X_test, y_test))\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "0f609176",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final accuracy: 92.1%\n"
     ]
    },
    {
     "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(accuracy_list[:])\n",
    "\n",
    "print(f\"Final accuracy: {100*accuracy_list[-8]:.1f}%\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}