第1 章AI 賦能時代的創(chuàng)業(yè)······················································································1
1.1 中國AI 如何彎道超車····································································································2
1.2 AI 從“發(fā)明期”進入“應用期”··················································································9
1.2.1 深度學習助推AI 進入“應用期”···································································10
1.2.2 To B 創(chuàng)業(yè)迎來黃金發(fā)展期···············································································.11
1.2.3 “傳統(tǒng)產業(yè)+AI”將創(chuàng)造巨大價值·····································································14
1.2.4 AI 賦能傳統(tǒng)行業(yè)四部曲···················································································16
1.3 AI 賦能時代的創(chuàng)業(yè)特點·······························································································21
1.3.1 海外科技巨頭成功因素解析·············································································21
1.3.2 科學家創(chuàng)業(yè)的優(yōu)勢和短板·················································································24
1.3.3 四因素降低AI 產品化、商業(yè)化門檻·······························································26
1.4 給未來AI 人才的建議··································································································30
第2 章AI 的產品化和工程化挑戰(zhàn)·········································································35
2.1 從AI 科研到AI 商業(yè)化································································································36
2.2 產品經理視角—數據驅動的產品研發(fā)······································································40
2.2.1 數據驅動············································································································41
2.2.2 典型C 端產品的設計和管理············································································43
2.2.3 典型B 端產品解決方案的設計和管理·····························································46
2.2.4 AI 技術的產品化·······························································································48
2.3 架構設計師視角—典型AI 架構···············································································51
2.3.1 為什么要重視系統(tǒng)架構····················································································51
2.3.2 與AI 相關的典型系統(tǒng)架構··············································································53
2.4 寫在本章最后的幾句話································································································78
本章參考文獻 ························································································································79
第3 章機器學習的發(fā)展現(xiàn)狀及前沿進展 ······························································81
3.1 機器學習的發(fā)展現(xiàn)狀····································································································82
3.2 機器學習的前沿進展····································································································85
3.2.1 復雜模型············································································································85
3.2.2 表示學習············································································································90
3.2.3 自動機器學習····································································································95
第4 章自然語言理解概述及主流任務 ··································································99
4.1 自然語言理解概述······································································································100
4.2 NLP 主流任務·············································································································100
4.2.1 中文分詞··········································································································101
4.2.2 指代消解··········································································································102
4.2.3 文本分類··········································································································103
4.2.4 關鍵詞（短語）的抽取與生成·······································································105
4.2.5 文本摘要··········································································································107
4.2.6 情感分析··········································································································108
本章參考文獻·····················································································································.111
第 5 章機器學習在 NLP 領域的應用及產業(yè)實踐···············································115
5.1 自然語言句法分析·····································································································.116
5.1.1 自然語言句法分析的含義與背景··································································.116
5.1.2 研究句法分析的幾個要素··············································································.117
5.1.3 句法分析模型舉例··························································································121
5.2 深度學習在句法分析模型參數估計中的應用····························································125
5.2.1 符號嵌入··········································································································126
5.2.2 上下文符號嵌入······························································································129
本章參考文獻······················································································································131
第 6 章計算機視覺前沿進展及實踐 ····································································133
6.1 計算機視覺概念··········································································································134
6.2 計算機視覺認知過程··································································································136
6.2.1 從低層次到高層次的理解···············································································137
6.2.2 基本任務及主流任務······················································································138
6.3 計算機視覺技術的前沿進展·······················································································141
6.3.1 圖像分類任務··································································································141
6.3.2 目標檢測任務··································································································148
6.3.3 圖像分割任務··································································································151
6.3.4 主流任務的前沿進展······················································································155
6.4 基于機器學習的計算機視覺實踐···············································································164
6.4.1 目標檢測比賽··································································································164
6.4.2 蛋筒質檢··········································································································167
6.4.3 智能貨柜··········································································································170
本章參考文獻······················································································································173
第 7 章深度學習模型壓縮與加速的技術發(fā)展與應用·········································175
7.1 深度學習的應用領域及面臨的挑戰(zhàn)···········································································176
7.1.1 深度學習的應用領域······················································································176
7.1.2 深度學習面臨的挑戰(zhàn)······················································································178
7.2 深度學習模型的壓縮和加速方法···············································································180
7.2.1 主流壓縮和加速方法概述···············································································180
7.2.2 權重剪枝··········································································································182
7.2.3 權重量化··········································································································192
7.2.4 知識蒸餾··········································································································199
7.2.5 權重量化與權重剪枝結合并泛化···································································200
7.3 模型壓縮與加速的應用場景·······················································································201
7.3.1 駕駛員安全檢測系統(tǒng)······················································································202
7.3.2 高級駕駛輔助系統(tǒng)··························································································202
7.3.3 車路協(xié)同系統(tǒng)··································································································203
本章參考文獻······················································································································204
第 8 章終端深度學習基礎、挑戰(zhàn)和工程實踐·····················································207
8.1 終端深度學習的技術成就及面臨的核心問題····························································208
8.1.1 終端深度學習的技術成就···············································································208
8.1.2 終端深度學習面臨的核心問題·······································································209
8.2 在冗余條件下減少資源需求的方法··········································································.211
8.3 在非冗余條件下減少資源需求的方法·······································································213
8.3.1 特殊化模型······································································································214
8.3.2 動態(tài)模型··········································································································215
8.4 深度學習系統(tǒng)的設計··································································································216
8.4.1 實際應用場景中的挑戰(zhàn)··················································································216
8.4.2 實際應用場景中的問題解決···········································································217
8.4.3 案例分析··········································································································219
本章參考文獻······················································································································224
第 9 章DeeCamp 訓練營最佳商業(yè)項目實戰(zhàn)·······················································225
9.1 方仔照相館—AI 輔助單張圖像生成積木方頭仔···················································227
9.1.1 讓“AI 方頭仔”觸手可及·············································································227
9.1.2 理論支撐：BiSeNet 和Mask R-CNN ·····························································229
9.1.3 任務分解：從圖像分析到積木生成的實現(xiàn)····················································231
9.1.4 團隊協(xié)作與時間安排······················································································237
9.2 AI 科幻世界—基于預訓練語言模型的科幻小說生成系統(tǒng)····································242
9.2.1 打造人機協(xié)作的科幻小說作家·······································································242
9.2.2 理論支撐：語言模型、Transformer 模型和GPT2 預訓練模型·····················243
9.2.3 從“找小說”到“寫小說”的實現(xiàn)步驟························································247
9.2.4 團隊協(xié)作與時間安排······················································································250
9.3 寵物健康識別—基于圖像表征學習的寵物肥胖度在線檢測系統(tǒng)·························254
9.3.1 人人都能做“養(yǎng)寵達人”···············································································254
9.3.2 理論支撐：表征學習、人臉識別原理和ArcFace 損失函數·························257
9.3.3 任務分解：從數據收集到肥胖度檢測···························································259
9.3.4 團隊協(xié)作與時間安排······················································································262
9.4 商品文案生成—基于檢索和生成的智能文案系統(tǒng)················································265
9.4.1 智能內容生成··································································································265
9.4.2 理論支撐：Word2Vec 詞嵌入、預訓練語言模型BERT 和Seq2Seq
文本生成··········································································································266
9.4.3 任務分解：“尋章摘句”和“文不加點”······················································269
9.4.4 團隊協(xié)作與時間安排······················································································273
本章參考文獻······················································································································276