第1 章 緒論····················································································.1
1.1 人工智能·············································································.2
1.1.1 人工智能技術(shù)的發(fā)展歷程···············································.3
1.1.2 人工智能技術(shù)的流派·····················································.9
1.2 深度學(xué)習(xí)與神經(jīng)網(wǎng)絡(luò)概述······················································.11
1.2.1 深度學(xué)習(xí)與神經(jīng)網(wǎng)絡(luò)技術(shù)的發(fā)展歷程······························.11
1.2.2 深度學(xué)習(xí)與神經(jīng)網(wǎng)絡(luò)的前沿技術(shù)····································.16
1.3 深度學(xué)習(xí)系統(tǒng)架構(gòu)·······························································.17
1.4 深度學(xué)習(xí)框架·····································································.19
1.5 深度學(xué)習(xí)的應(yīng)用··································································.20
1.5.1 計(jì)算機(jī)視覺(jué)·······························································.20
1.5.2 語(yǔ)音語(yǔ)義··································································.21
1.5.3 自然語(yǔ)言處理····························································.22
1.6 人工智能潛在的安全風(fēng)險(xiǎn)······················································.22
1.6.1 數(shù)據(jù)層面的風(fēng)險(xiǎn)·························································.23
1.6.2 算法模型層面的風(fēng)險(xiǎn)···················································.23
1.6.3 智能計(jì)算框架層面的風(fēng)險(xiǎn)·············································.23
1.6.4 基礎(chǔ)軟硬件層面的風(fēng)險(xiǎn)················································.24
1.6.5 應(yīng)用服務(wù)層面的風(fēng)險(xiǎn)···················································.24
本章小結(jié)··················································································.24
第2 章 預(yù)備知識(shí)············································································.25
2.1 相關(guān)數(shù)學(xué)基礎(chǔ)·····································································.25
2.1.1 線(xiàn)性代數(shù)··································································.25
2.1.2 概率論·····································································.27
2.1.3 優(yōu)化理論··································································.32
2.2 機(jī)器學(xué)習(xí)基礎(chǔ)·····································································.34
2.2.1 機(jī)器學(xué)習(xí)算法的基本流程·············································.35
2.2.2 機(jī)器學(xué)習(xí)常用評(píng)價(jià)指標(biāo)················································.36
2.2.3 典型機(jī)器學(xué)習(xí)算法······················································.41
2.3 實(shí)驗(yàn)環(huán)境基礎(chǔ)·····································································.49
2.3.1 GPU 驅(qū)動(dòng)的安裝配置··················································.49
2.3.2 依賴(lài)環(huán)境的安裝配置···················································.52
2.3.3 深度學(xué)習(xí)框架的安裝配置·············································.54
2.3.4 集成開(kāi)發(fā)環(huán)境的安裝配置·············································.56
本章小結(jié)··················································································.57
第3 章 前饋神經(jīng)網(wǎng)絡(luò)······································································.58
3.1 感知器··············································································.58
3.1.1 單層感知器·······························································.58
3.1.2 多層感知器·······························································.61
3.1.3 前饋神經(jīng)網(wǎng)絡(luò)的基本結(jié)構(gòu)·············································.62
3.2 激活函數(shù)···········································································.63
3.2.1 Sigmoid 函數(shù)·····························································.63
3.2.2 ReLU 函數(shù)································································.65
3.2.3 Tanh 函數(shù)·································································.67
3.2.4 Softmax 函數(shù)·····························································.68
3.3 誤差反向傳播·····································································.69
3.3.1 梯度下降法·······························································.69
3.3.2 鏈?zhǔn)椒▌t··································································.70
3.3.3 反向傳播··································································.72
本章小結(jié)··················································································.75
第4 章 深度模型的優(yōu)化···································································.76
4.1 神經(jīng)網(wǎng)絡(luò)的優(yōu)化問(wèn)題····························································.76
4.1.1 局部最優(yōu)和振蕩陷阱···················································.76
4.1.2 梯度爆炸和梯度消失···················································.78
4.2 常見(jiàn)的優(yōu)化算法··································································.80
4.2.1 梯度下降優(yōu)化算法······················································.80
4.2.2 二階優(yōu)化算法····························································.81
4.3 自適應(yīng)學(xué)習(xí)率算法·······························································.85
4.3.1 AdaGrad 算法····························································.85
4.3.2 RMSprop 算法···························································.86
4.3.3 Adam 算法································································.86
4.4 參數(shù)初始化方法··································································.87
4.4.1 隨機(jī)初始化·······························································.87
4.4.2 Xavier 初始化····························································.88
4.4.3 He 初始化·································································.89
本章小結(jié)··················································································.90
第5 章 深度學(xué)習(xí)中的正則化·····························································.91
5.1 范數(shù)懲罰···········································································.91
5.1.1 L1 正則化··································································.91
5.1.2 L2 正則化··································································.93
5.2 數(shù)據(jù)集增強(qiáng)與噪聲注入·························································.94
5.2.1 數(shù)據(jù)集增強(qiáng)·······························································.94
5.2.2 噪聲注入··································································.95
5.3 提前停止···········································································.97
5.4 Dropout ·············································································.97
5.5 批歸一化···········································································.99
本章小結(jié)·················································································.102
第6 章 卷積神經(jīng)網(wǎng)絡(luò)······································································103
6.1 卷積神經(jīng)網(wǎng)絡(luò)的發(fā)展歷程·····················································.103
6.2 卷積神經(jīng)網(wǎng)絡(luò)的基本組成·····················································.104
6.2.1 卷積層····································································.105
6.2.2 池化層····································································.106
6.2.3 全連接層·································································.108
6.3 常見(jiàn)卷積神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)························································.108
6.3.1 VGG 網(wǎng)絡(luò)································································.109
6.3.2 GoogLeNet 網(wǎng)絡(luò)························································.111
6.3.3 ResNet 網(wǎng)絡(luò)·····························································.117
6.4 深度生成網(wǎng)絡(luò)····································································.122
6.4.1 生成對(duì)抗網(wǎng)絡(luò)···························································.122
6.4.2 深度卷積生成對(duì)抗網(wǎng)絡(luò)···············································.123
6.5 圖像分類(lèi)案例····································································.134
6.5.1 步驟1：搭建環(huán)境······················································.135
6.5.2 步驟2：導(dǎo)入依賴(lài)庫(kù)···················································.137
6.5.3 步驟3：獲取數(shù)據(jù)······················································.137
6.5.4 步驟4：定義AlexNet 網(wǎng)絡(luò)··········································.138
6.5.5 步驟5：模型初始化···················································.140
6.5.6 步驟6：模型訓(xùn)練······················································.140
6.6 目標(biāo)檢測(cè)案例····································································.143
6.6.1 步驟1：環(huán)境配置和模型下載·······································.144
6.6.2 步驟2：主函數(shù)解析···················································.151
6.6.3 步驟3：終端指令運(yùn)行················································.152
本章小結(jié)·················································································.153
第7 章 循環(huán)神經(jīng)網(wǎng)絡(luò)·····································································.154
7.1 循環(huán)神經(jīng)網(wǎng)絡(luò)的基本原理·····················································.154
7.1.1 循環(huán)神經(jīng)網(wǎng)絡(luò)的原理··················································.154
7.1.2 雙向循環(huán)神經(jīng)網(wǎng)絡(luò)·····················································.157
7.2 循環(huán)神經(jīng)網(wǎng)絡(luò)在實(shí)際中的應(yīng)用···············································.158
7.2.1 文本生成·································································.159
7.2.2 語(yǔ)音識(shí)別·································································.159
7.2.3 機(jī)器翻譯·································································.160
7.2.4 生成圖像描述···························································.161
7.2.5 視頻動(dòng)作檢測(cè)···························································.162
7.2.6 信號(hào)分類(lèi)·································································.162
7.3 長(zhǎng)短期記憶網(wǎng)絡(luò)及其他門(mén)控循環(huán)神經(jīng)網(wǎng)絡(luò)································.163
7.3.1 長(zhǎng)短期記憶網(wǎng)絡(luò)························································.164
7.3.2 其他門(mén)控循環(huán)神經(jīng)網(wǎng)絡(luò)···············································.166
7.4 深度學(xué)習(xí)在文本和序列中的應(yīng)用············································.167
7.4.1 文本數(shù)據(jù)處理···························································.167
7.4.2 文本分類(lèi)和情感分析··················································.180
7.4.3 機(jī)器翻譯·································································.180
7.4.4 命名實(shí)體識(shí)別···························································.182
7.5 卷積神經(jīng)網(wǎng)絡(luò)與循環(huán)神經(jīng)網(wǎng)絡(luò)···············································.183
7.5.1 卷積神經(jīng)網(wǎng)絡(luò)與循環(huán)神經(jīng)網(wǎng)絡(luò)的對(duì)比·····························.183
7.5.2 卷積神經(jīng)網(wǎng)絡(luò)與循環(huán)神經(jīng)網(wǎng)絡(luò)的組合應(yīng)用·······················.184
7.6 案例：深度學(xué)習(xí)的詩(shī)歌生成··················································.185
7.6.1 步驟1：導(dǎo)入依賴(lài)庫(kù)···················································.186
7.6.2 步驟2：讀取數(shù)據(jù)······················································.187
7.6.3 步驟3：構(gòu)造數(shù)據(jù)集···················································.188
7.6.4 步驟4：構(gòu)造模型······················································.190
7.6.5 步驟5：訓(xùn)練過(guò)程······················································.192
7.6.6 步驟6：生成文本······················································.195
7.7 案例：基于LSTM 算法的股票預(yù)測(cè)·········································.196
7.7.1 步驟1：導(dǎo)入依賴(lài)庫(kù)···················································.196
7.7.2 步驟2：獲取并處理數(shù)據(jù)·············································.197
7.7.3 步驟3：構(gòu)建預(yù)測(cè)數(shù)據(jù)序列··········································.199
7.7.4 步驟4：構(gòu)建LSTM 網(wǎng)絡(luò)············································.201
7.7.5 步驟5：訓(xùn)練網(wǎng)絡(luò)······················································.202
7.7.6 步驟6：預(yù)測(cè)測(cè)試集···················································.203
7.8 案例：基于深度學(xué)習(xí)的文本分類(lèi)············································.204
7.8.1 步驟1：項(xiàng)目入口······················································.205
7.8.2 步驟2：訓(xùn)練模塊······················································.207
7.8.3 步驟3：驗(yàn)證和測(cè)試函數(shù)·············································.210
7.8.4 步驟4：數(shù)據(jù)預(yù)處理模塊·············································.211
7.8.5 步驟5：定義模型······················································.215
7.8.6 步驟6：分類(lèi)結(jié)果展示················································.218
本章小結(jié)·················································································.219