liear_reg.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d76f5799-0baa-41f7-8be1-221d6fb8e59b",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": "skip"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import numpy as np \n",
    "import matplotlib.pyplot as plt \n",
    "import pandas as pd \n",
    "import jax.numpy as jnp\n",
    "from jax import grad, jit, vmap\n",
    "from jax import random\n",
    "\n",
    "key = random.PRNGKey(0)  # 使用任意整数作为种子\n",
    "key, subkey = random.split(key)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "b8130f3c-64be-44df-bc7f-e5abff9ae890",
   "metadata": {
    "editable": true,
    "slideshow": {
     "slide_type": "skip"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "w0 is: \n",
      "[[-2.4424558  -2.0356805   0.20554423]\n",
      " [-0.3535502  -0.76197404 -1.1785518 ]\n",
      " [-1.1482196   0.29716578 -1.3105359 ]\n",
      " [ 2.1302025  -0.18957235  0.96401215]\n",
      " [-1.3011001  -0.7486938  -0.3729984 ]]\n",
      "b0 is: [-2.4424558]\n"
     ]
    }
   ],
   "source": [
    "dim = (5, 3)\n",
    "\n",
    "X = random.normal(subkey, (10000, 5))\n",
    "w0 = random.normal(subkey, dim)\n",
    "b0 = random.normal(subkey, 1)\n",
    "print(f'w0 is: \\n{w0}')\n",
    "print(f'b0 is: {b0}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "defa542a-b81f-40f5-a32a-91d08bd17951",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 3)\n"
     ]
    }
   ],
   "source": [
    "y = X @ w0 + b0\n",
    "y += random.normal(subkey, (10000, y.shape[1])) * 0.01\n",
    "\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5feeda1b-c3b2-4b77-ac61-af06e70064ec",
   "metadata": {},
   "outputs": [],
   "source": [
    "lo = lambda y, y_pred: jnp.mean((y - y_pred)**2)\n",
    "loss = lambda w, b: lo(y, X @ w + b)\n",
    "loss = jit(loss)  # time: 3.3542935848236084 -> 1.253995656967163 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "cecd0beb-8395-4d8d-9d0f-26e93a8e048b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[*] time cost while training: 0.26192569732666016\n",
      "[*] w after training: \n",
      "[[-2.4425545  -2.0356748   0.20547928]\n",
      " [-0.3534928  -0.7620101  -1.1786641 ]\n",
      " [-1.148321    0.29706654 -1.3105948 ]\n",
      " [ 2.1301367  -0.1895199   0.96406233]\n",
      " [-1.3011376  -0.74883085 -0.3731989 ]]\n",
      "[*] b after training: [-2.442409]\n"
     ]
    }
   ],
   "source": [
    "w, b = random.normal(subkey, dim), random.normal(subkey, 1)\n",
    "\n",
    "alpha = 0.1\n",
    "ll = []\n",
    "import time\n",
    "start = time.time()\n",
    "for _ in range(100):\n",
    "    ll.append(loss(w, b))\n",
    "    dw = -grad(loss, argnums=0)(w, b)\n",
    "    db = -grad(loss, argnums=1)(w, b)\n",
    "    w += alpha * dw\n",
    "    b += alpha * db\n",
    "\n",
    "print(f'[*] time cost while training: {time.time() - start}')\n",
    "\n",
    "print(f'[*] w after training: \\n{w}')\n",
    "print(f'[*] b after training: {b}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "68249238-8a4c-41c0-956a-edaab73eca2b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(range(len(ll)), ll, label='loss with epochs')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.13.2+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}