●目 錄深度學習基礎篇 章 緒論 ·································································································· 21.1 引言 ······································································································· 21.1.1 Google 的深度學習成果 ···························································· 21.1.2 Microsoft 的深度學習成果························································· 31.1.3 國內公司的深度學習成果 ························································· 31.2 深度學習技術的發展歷程 ···································································· 41.3 深度學習的應用領域 ············································································ 61.3.1 圖像識別領域 ············································································· 61.3.2 語音識別領域 ············································································· 61.3.3 自然語言理解領域 ····································································· 71.4 如何開展深度學習的研究和應用開發 ················································· 7本章參考文獻 ······························································································ 11第2 章 國內外深度學習技術研發現狀及其產業化趨勢 ······························· 132.1 Google 在深度學習領域的研發現狀 ·················································· 132.1.1 深度學習在Google 的應用 ······················································ 132.1.2 Google 的TensorFlow 深度學習平臺 ······································ 142.1.3 Google 的深度學習芯片TPU ·················································· 152.2 Facebook 在深度學習領域的研發現狀 ·············································· 152.2.1 Torchnet ···················································································· 152.2.2 DeepText ··················································································· 162.3 百度在深度學習領域的研發現狀 ······················································· 172.3.1 光學字符識別 ··········································································· 172.3.2 商品圖像搜索 ··········································································· 172.3.3 在線廣告 ·················································································· 182.3.4 以圖搜圖 ·················································································· 182.3.5 語音識別 ·················································································· 182.3.6 百度開源深度學習平臺MXNet 及其改進的深度語音識別繫統Warp-CTC ····· 192.4 阿裡巴巴在深度學習領域的研發現狀 ··············································· 192.4.1 拍立淘 ······················································································ 192.4.2 阿裡小蜜――智能客服Messenger ········································· 202.5 京東在深度學習領域的研發現狀 ······················································· 202.6 騰訊在深度學習領域的研發現狀 ······················································· 212.7 科創型公司（基於深度學習的人臉識別繫統） ······························· 222.8 深度學習的硬件支撐――NVIDIA GPU ············································ 23本章參考文獻 ······························································································ 24深度學習理論篇第3 章 神經網絡 ························································································· 303.1的概念 ······················································································ 303.2 神經網絡 ····························································································· 313.2.1 後向傳播算法 ··········································································· 323.2.2 後向傳播算法推導 ··································································· 333.3 神經網絡算法示例 ·············································································· 36本章參考文獻 ······························································································ 38第4 章 卷積神經網絡 ················································································· 394.1 卷積神經網絡特性 ················································································ 394.1.1 局部連接 ·················································································· 404.1.2 權值共享 ·················································································· 414.1.3 空間相關下采樣 ······································································· 424.2 卷積神經網絡操作 ·············································································· 424.2.1 卷積操作 ·················································································· 424.2.2 下采樣操作 ·············································································· 444.3 卷積神經網絡示例：LeNet-5 ····························································· 45本章參考文獻 ······························································································ 48深度學習工具篇第5 章 深度學習工具Caffe ········································································ 505.1 Caffe 的安裝 ························································································ 505.1.1 安裝依賴包 ·············································································· 515.1.2 CUDA 安裝 ·············································································· 515.1.3 MATLAB 和Python 安裝 ························································ 545.1.4 OpenCV 安裝（可選） ···························································· 595.1.5 Intel MKL 或者BLAS 安裝 ····················································· 595.1.6 Caffe 編譯和測試 ····································································· 595.1.7 Caffe 安裝問題分析 ································································· 625.2 Caffe 框架與源代碼解析 ···································································· 635.2.1 數據層解析 ·············································································· 635.2.2 網絡層解析 ·············································································· 745.2.3 網絡結構解析 ··········································································· 925.2.4 網絡求解解析 ········································································· 104本章參考文獻 ···························································································· 109第6 章 深度學習工具Pylearn2 ································································ 1106.1 Pylearn2 的安裝 ·················································································· 1106.1.1 相關依賴安裝 ·········································································· 1106.1.2 安裝Pylearn2 ·········································································· 1126.2 Pylearn2 的使用 ·················································································· 112本章參考文獻 ····························································································· 116深度學習實踐篇（入門與進階）第7 章 基於深度學習的手寫數字識別 ······················································ 1187.1 數據介紹 ···························································································· 1187.1.1 MNIST 數據集 ········································································ 1187.1.2 提取MNIST 數據集圖片 ······················································· 1207.2 手寫字體識別流程 ············································································ 1217.2.1 模型介紹 ················································································ 1217.2.2 操作流程 ················································································ 1267.3 實驗結果分析 ···················································································· 127本章參考文獻 ···························································································· 128第8 章 基於深度學習的圖像識別 ····························································· 1298.1 數據來源 ··························································································· 1298.1.1 Cifar10 數據集介紹 ································································ 1298.1.2 Cifar10 數據集格式 ································································ 1298.2 Cifar10 識別流程 ··············································································· 1308.2.1 模型介紹 ················································································ 1308.2.2 操作流程 ················································································ 1368.3 實驗結果分析 ······················································································ 139本章參考文獻 ···························································································· 140第9 章 基於深度學習的物體圖像識別 ······················································ 1419.1 數據來源 ··························································································· 1419.1.1 Caltech101 數據集 ·································································· 1419.1.2 Caltech101 數據集處理 ·························································· 1429.2 物體圖像識別流程 ············································································ 1439.2.1 模型介紹 ················································································ 1439.2.2 操作流程 ················································································ 1449.3 實驗結果分析 ···················································································· 150本章參考文獻 ···························································································· 1510 章 基於深度學習的人臉識別 ··························································· 15210.1 數據來源 ························································································· 15210.1.1 AT＆T Facedatabase 數據庫 ·················································· 15210.1.2 數據庫處理 ··········································································· 15210.2 人臉識別流程 ·················································································· 15410.2.1 模型介紹 ·············································································· 15410.2.2 操作流程 ·············································································· 15510.3 實驗結果分析 ·················································································· 159本章參考文獻 ···························································································· 160深度學習實踐篇（高級應用）1 章 基於深度學習的人臉識別――DeepID 算法 ································ 16211.1 問題定義與數據來源 ······································································ 16211.2 算法原理 ·························································································· 16311.2.1 數據預處理 ··········································································· 16311.2.2 模型訓練策略 ······································································· 16411.2.3 算法驗證和結果評估 ··························································· 16411.3 人臉識別步驟 ·················································································· 16511.3.1 數據預處理 ··········································································· 16511.3.2 深度網絡結構模型 ······························································· 16811.3.3 提取深度特征與人臉驗證 ··················································· 17111.4 實驗結果分析 ·················································································· 17411.4.1 實驗數據 ··············································································· 17411.4.2 實驗結果分析 ······································································· 175本章參考文獻 ···························································································· 1762 章 基於深度學習的表情識別 ··························································· 17712.1 表情數據 ························································································· 17712.1.1 Cohn-Kanade（CK+）數據庫 ············································· 17712.1.2 JAFFE 數據庫 ······································································ 17812.2 算法原理 ························································································· 17912.3 表情識別步驟 ·················································································· 18012.3.1 數據預處理 ··········································································· 18012.3.2 深度神經網絡結構模型 ······················································· 18112.3.3 提取深度特征及分類 ··························································· 18212.4 實驗結果分析 ·················································································· 18412.4.1 實現細節 ·············································································· 18412.4.2 實驗結果對比 ······································································· 185本章參考文獻 ···························································································· 1883 章 基於深度學習的年齡估計 ··························································· 19013.1 問題定義 ························································································· 19013.2 年齡估計算法 ·················································································· 19013.2.1 數據預處理 ··········································································· 19013.2.2 提取深度特征 ······································································· 19213.2.3 提取LBP 特征 ····································································· 19613.2.4 訓練回歸模型 ······································································· 19613.3 實驗結果分析 ·················································································· 199本章參考文獻 ···························································································· 1994 章 基於深度學習的人臉關鍵點檢測 ················································ 20014.1 問題定義和數據來源 ······································································ 20014.2 基於深度學習的人臉關鍵點檢測的步驟 ······································· 20114.2.1 數據預處理 ··········································································· 20114.2.2 訓練深度學習網絡模型 ······················································· 20614.2.3 預測和處理關鍵點坐標 ······················································· 207本章參考文獻 ···························································································· 212深度學結與展望篇5 章 總結與展望 ················································································· 21415.1 深度學習領域當前的主流技術及其應用領域 ······························· 21415.1.1 圖像識別 ·············································································· 21415.1.2 語音識別與自然語言理解 ··················································· 21515.2 深度學習的缺陷 ·············································································· 21515.2.1 深度學習在硬件方面的門檻較高 ········································ 21515.2.2 深度學習在軟件安裝與配置方面的門檻較高 ···················· 21615.2.3 深度學習最重要的問題在於需要海量的有標注的數據作為支撐 ··· 21615.2.4 深度學習的最後階段竟然變成枯燥、機械、及其耗時的調參工作 ··· 21715.2.5 深度學習不適用於數據量較小的數據 ································ 21815.2.6 深度學習目前主要用於圖像、聲音的識別和自然語言的理解 ····· 21815.2.7 研究人員從事深度學習研究的困境 ···································· 21915.3 展望 ································································································· 220本章參考文獻 ···························································································· 220

本書全面、繫統地介紹深度學習相關的技術，包括人工神經網絡，卷積神經網絡，深度學習平臺及源代碼分析，深度學習入門與進階，深度學習不錯實踐，所有章節均附有源程序，所有實驗讀者均可重現，具有高度的可操作性和實用性。通過學習本書，研究人員、深度學習愛好者，能夠在3 個月內，繫統掌握深度學習相關的理論和技術。

張重生 著