03-Softmax_Classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 多分类问题最后一层经过Softmax Layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src='picture/softmax1.jpg', width=600>\n",
       "<img src='picture/softmax2.jpg', width=600>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%html\n",
    "<img src='picture/softmax1.jpg', width=600>\n",
    "<img src='picture/softmax2.jpg', width=600>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 代码实现，3分类为例"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src='picture/jisuan1.jpg', width=600>\n",
       "<img src='picture/jisuan2.jpg', width=600>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%html\n",
    "<img src='picture/jisuan1.jpg', width=600>\n",
    "<img src='picture/jisuan2.jpg', width=600>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### CrossEntropyLoss <==> LogSoftmax + NLLLoss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src='picture/NLLloss.jpg', width=600>\n",
       "<img src='picture/Cross.jpg', width=600>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%html\n",
    "<img src='picture/NLLloss.jpg', width=600>\n",
    "<img src='picture/Cross.jpg', width=600>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "criterion = torch.nn.CrossEntropyLoss()\n",
    "Y = torch.LongTensor([2, 0, 1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_pred1 = torch.Tensor([[0.1, 0.2, 0.9],\n",
    "                        [1.1, 0.1, 0.2],\n",
    "                        [0.2, 2.1, 0.1]])\n",
    "Y_pred2 = torch.Tensor([[0.8, 0.2, 0.3],\n",
    "                        [0.2, 0.3, 0.5],\n",
    "                        [0.2, 0.2, 0.5]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Batch loss1 =  tensor(0.4966) \n",
      "Batch Loss2 =  tensor(1.2389)\n"
     ]
    }
   ],
   "source": [
    "l1 = criterion(Y_pred1, Y)\n",
    "l2 = criterion(Y_pred2, Y)\n",
    "print(\"Batch loss1 = \", l1.data, \"\\nBatch Loss2 = \", l2.data)"
   ]
  }
 ],
 "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}