{ "cells": [ { "cell_type": "markdown", "metadata": { "slideshow": { "slide_type": "slide" } }, "source": [ "# 「深層学習」読書会 〜第6章〜" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "
2016/05/21 機械学習 名古屋 第4回勉強会
" ] }, { "cell_type": "markdown", "metadata": { "slideshow": { "slide_type": "slide" } }, "source": [ "## 第6章 畳込みニューラルネット" ] }, { "cell_type": "markdown", "metadata": { "slideshow": { "slide_type": "slide" } }, "source": [ "### 6.1 単純型細胞と複雑型細胞" ] }, { "cell_type": "markdown", "metadata": { "slideshow": { "slide_type": "notes" } }, "source": [ "後回し。 \n", "《TODO: やるとしても言葉の説明だけ?》 \n", "《TODO: 図を描くか?》" ] }, { "cell_type": "markdown", "metadata": { "slideshow": { "slide_type": "slide" } }, "source": [ "### 6.2 全体の構造" ] }, { "cell_type": "code", "execution_count": 1, "metadata": { "collapsed": true, "slideshow": { "slide_type": "skip" } }, "outputs": [], "source": [ "from graphviz import Digraph" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "collapsed": true, "slideshow": { "slide_type": "skip" } }, "outputs": [], "source": [ "fig_6_3 = Digraph(\"fig_6_3\", format=\"svg\")\n", "\n", "fig_6_3.body.extend(['rankdir=LR'])\n", "\n", "c1 = Digraph('cluster_1')\n", "c1.body.append('style=filled')\n", "c1.body.append('color=white')\n", "c1.node_attr.update(shape='box', style='rounded,filled', color='white', penwidth='2')\n", "c1.attr('edge', color=\"#737373\")\n", "c1.edges([('input', 'conv0'), ('conv0', 'conv1'), ('conv1', 'pooling1'), ('pooling1', 'lcm1')])\n", "c1.node('input', 'input (image)', fillcolor=\"#b3ffff\", color=\"#009999\")\n", "c1.node('conv0', 'convolution', fillcolor=\"#fb8072\", color=\"#941305\")\n", "c1.node('conv1', 'convolution', fillcolor=\"#fb8072\", color=\"#941305\")\n", "c1.node('pooling1', 'pooling', fillcolor=\"#80b1d3\", color=\"#275372\")\n", "c1.node('lcm1', 'LCM', fillcolor=\"#eeeeee\", color=\"#808080\")\n", "\n", "c2 = Digraph('cluster_2')\n", "c2.body.append('style=filled')\n", "c2.body.append('color=white')\n", "c2.node_attr.update(shape='box', style='rounded,filled', color='white', penwidth='2')\n", "c2.attr('edge', color=\"#737373\")\n", "c2.edges([('conv2', 'pooling2'), ('pooling2', 'fc1'), ('fc1', 'fc2'), ('fc2', 'softmax'), ('softmax', 'output')])\n", "c2.node('conv2', 'convolution', fillcolor=\"#fb8072\", color=\"#941305\")\n", "c2.node('pooling2', 'pooling', fillcolor=\"#80b1d3\", color=\"#275372\")\n", "c2.node('fc1', 'fully-\\nconnected', fillcolor=\"#ffffb3\", color=\"#999900\")\n", "c2.node('fc2', 'fully-\\nconnected', fillcolor=\"#ffffb3\", color=\"#999900\")\n", "c2.node('softmax', 'softmax', fillcolor=\"#b3de69\", color=\"#597d1c\")\n", "c2.node('output', 'output (label)', fillcolor=\"#b3ffff\", color=\"#009999\")\n", "\n", "fig_6_3.subgraph(c1)\n", "fig_6_3.subgraph(c2)\n", "\n", "fig_6_3.attr('edge', color=\"#737373\")\n", "\n", "fig_6_3.edge('conv2', 'lcm1', dir=\"back\", minlen=\"4\")\n" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "collapsed": false, "slideshow": { "slide_type": "subslide" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n", "\n" ], "text/plain": [ "