目 錄
第1 章 蒙特卡羅方法簡介··············································································.1
1.1 引言·······························································································.1
1.2 動機(jī)和目標(biāo)······················································································.1
1.3 蒙特卡羅計算中的任務(wù)·······································································.2
1.3.1 任務(wù)1：采樣和模擬········································································.3
1.3.2 任務(wù)2：通過蒙特卡羅模擬估算未知量···················································.5
1.3.3 任務(wù)3：優(yōu)化和貝葉斯推理································································.7
1.3.4 任務(wù)4：學(xué)習(xí)和模型估計···································································.8
1.3.5 任務(wù)5：可視化能級圖·····································································.9
本章參考文獻(xiàn)··························································································13
第2 章 序貫蒙特卡羅方法··············································································14
2.1 引言·······························································································14
2.2 一維密度采樣···················································································14
2.3 重要性采樣和加權(quán)樣本·······································································15
2.4 序貫重要性采樣（SIS） ······································································18
2.4.1 應(yīng)用：表達(dá)聚合物生長的自避游走························································18
2.4.2 應(yīng)用：目標(biāo)跟蹤的非線性/粒子濾波·······················································20
2.4.3 SMC 方法框架總結(jié)·········································································23
2.5 應(yīng)用：利用SMC 方法進(jìn)行光線追蹤·······················································24
2.6 在重要性采樣中保持樣本多樣性···························································25
2.6.1 基本方法····················································································25
2.6.2 Parzen 窗討論··············································································28
2.7 蒙特卡羅樹搜索················································································29
2.7.1 純蒙特卡羅樹搜索··········································································30
2.7.2 AlphaGo ·····················································································32
2.8 本章練習(xí)·························································································33
本章參考文獻(xiàn)··························································································35
第3 章 馬爾可夫鏈蒙特卡羅方法基礎(chǔ)·······························································36
3.1 引言·······························································································36
蒙特卡羅方法與人工智能
·X ·
3.2 馬爾可夫鏈基礎(chǔ)················································································37
3.3 轉(zhuǎn)移矩陣的拓?fù)洌哼B通與周期······························································38
3.4 Perron-Frobenius 定理··········································································41
3.5 收斂性度量······················································································42
3.6 連續(xù)或異構(gòu)狀態(tài)空間中的馬爾可夫鏈·····················································44
3.7 各態(tài)遍歷性定理················································································45
3.8 通過模擬退火進(jìn)行MCMC 優(yōu)化·····························································46
3.9 本章練習(xí)·························································································49
本章參考文獻(xiàn)··························································································51
第4 章 Metropolis 算法及其變體······································································52
4.1 引言·······························································································52
4.2 Metropolis-Hastings 算法······································································52
4.2.1 原始Metropolis-Hastings 算法······························································53
4.2.2 Metropolis-Hastings 算法的另一形式·······················································54
4.2.3 其他接受概率設(shè)計··········································································55
4.2.4 Metropolis 算法設(shè)計中的關(guān)鍵問題·························································55
4.3 獨立Metropolis 采樣···········································································55
4.3.1 IMS 的特征結(jié)構(gòu)············································································56
4.3.2 有限空間的一般首中時·····································································57
4.3.3 IMS 擊中時分析············································································57
4.4 可逆跳躍和跨維MCMC ······································································59
4.4.1 可逆跳躍····················································································59
4.4.2 簡單例子：一維圖像分割··································································60
4.5 應(yīng)用：計算人數(shù)················································································63
4.5.1 標(biāo)值點過程模型············································································64
4.5.2 MCMC 推理·················································································64
4.5.3 結(jié)果·························································································65
4.6 應(yīng)用：家具布置················································································65
4.7 應(yīng)用：場景合成················································································67
4.8 本章練習(xí)·························································································71
本章參考文獻(xiàn)··························································································72
第5 章 吉布斯采樣器及其變體········································································73
5.1 引言·······························································································73
5.2 吉布斯采樣器···················································································74
目 錄
·XI·
5.2.1 吉布斯采樣器介紹··········································································74
5.2.2 吉布斯采樣器的一個主要問題·····························································75
5.3 吉布斯采樣器擴(kuò)展·············································································76
5.3.1 擊中逃跑····················································································77
5.3.2 廣義吉布斯采樣器··········································································77
5.3.3 廣義擊中逃跑···············································································77
5.3.4 利用輔助變量采樣··········································································78
5.3.5 模擬退火····················································································78
5.3.6 切片采樣····················································································79
5.3.7 數(shù)據(jù)增強(qiáng)····················································································80
5.3.8 Metropolized 吉布斯采樣器·································································80
5.4 數(shù)據(jù)關(guān)聯(lián)和數(shù)據(jù)增強(qiáng)··········································································82
5.5 Julesz 系綜和MCMC 紋理采樣······························································83
5.5.1 Julesz 系綜：紋理的數(shù)學(xué)定義······························································84
5.5.2 吉布斯系綜和系綜等價性··································································85
5.5.3 Julesz 系綜采樣·············································································86
5.5.4 實驗：對Julesz 系綜進(jìn)行采樣·····························································87
5.6 本章練習(xí)·························································································89
本章參考文獻(xiàn)··························································································90
第6 章 聚類采樣方法····················································································91
6.1 引言·······························································································91
6.2 Potts 模型和SW 算法·········································································92
6.3 SW 算法詳解····················································································94
6.3.1 解釋1：Metropolis-Hastings 觀點··························································94
6.3.2 解釋2：數(shù)據(jù)增強(qiáng)··········································································97
6.4 SW 算法的相關(guān)理論結(jié)果··································································.100
6.5 任意概率的SW 切分算法·································································.102
6.5.1 步驟一：數(shù)據(jù)驅(qū)動的聚類·······························································.102
6.5.2 步驟二：顏色翻轉(zhuǎn)·······································································.103
6.5.3 步驟三：接受翻轉(zhuǎn)·······································································.104
6.5.4 復(fù)雜性分析···············································································.105
6.6 聚類采樣方法的變體·······································································.106
6.6.1 聚類吉布斯采樣：“擊中逃跑”觀點·····················································.106
6.6.2 多重翻轉(zhuǎn)方案············································································.107
6.7 應(yīng)用：圖像分割·············································································.107
蒙特卡羅方法與人工智能
·X II·
6.8 多重網(wǎng)格和多級SW 切分算法···························································.110
6.8.1 多重網(wǎng)格SW 切分算法··································································.111
6.8.2 多級SW 切分算法·······································································.113
6.9 子空間聚類···················································································.114
6.9.1 通過SW 切分算法進(jìn)行子空間聚類·····················································.115
6.9.2 應(yīng)用：稀疏運(yùn)動分割····································································.117
6.10 C 4：聚類合作競爭約束··································································.121
6.10.1 C 4 算法綜述············································································.123
6.10.2 圖形、耦合和聚類······································································.124
6.10.3 平面圖上的C 4 算法····································································.128
6.10.4 在平面圖上的實驗······································································.131
6.10.5 棋盤Ising 模型·········································································.132
6.10.6 分層圖上的C 4··········································································.136
6.10.7 C 4 分層實驗············································································.138
6.11 本章練習(xí)·····················································································.139
本章參考文獻(xiàn)·······················································································.140
第7 章 MCMC 的收斂性分析·······································································.144
7.1 引言····························································································.144
7.2 關(guān)鍵收斂問題················································································.144
7.3 實用的監(jiān)測方法·············································································.145
7.4 洗牌的耦合方法·············································································.146
7.4.1 置頂洗牌·················································································.147
7.4.2 Riffle 洗牌················································································.147
7.5 幾何界限、瓶頸和連通率·································································.149
7.5.1 幾何收斂·················································································.149
7.5.2 交易圖（轉(zhuǎn)換圖）·······································································.150
7.5.3 瓶頸······················································································.150
7.5.4 連通率····················································································.151
7.6 Peskun 有序和遍歷性定理·································································.152
7.7 路徑耦合和精確采樣·······································································.153
7.7.1 從過去耦合···············································································.154
7.7.2 應(yīng)用：對Ising 模型進(jìn)行采樣···························································.155
7.8 本章練習(xí)······················································································.157
本章參考文獻(xiàn)·······················································································.159
目 錄
·XIII·
第8 章 數(shù)據(jù)驅(qū)動的馬爾可夫鏈蒙特卡羅方法···················································.160
8.1 引言····························································································.160
8.2 圖像分割和DDMCMC 方法概述························································.160
8.3 DDMCMC 方法解釋········································································.161
8.3.1 MCMC 方法設(shè)計的基本問題····························································.163
8.3.2 計算原子空間中的提議概率：原子粒子················································.164
8.3.3 計算對象空間中的提議概率：對象粒子················································.166
8.3.4 計算多個不同的解：場景粒子··························································.167
8.3.5 Ψ-世界實驗··············································································.167
8.4 問題表達(dá)和圖像建�！ぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁ�.168
8.4.1 用于分割的貝葉斯公式··································································.169
8.4.2 先驗概率·················································································.169
8.4.3 灰度圖像的似然·········································································.169
8.4.4 模型校準(zhǔn)·················································································.171
8.4.5 彩色圖像模型············································································.172
8.5 解空間分析···················································································.173
8.6 使用遍歷馬爾可夫鏈探索解空間························································.174
8.6.1 五類馬爾可夫鏈動態(tài)過程·······························································.174
8.6.2 瓶頸問題·················································································.175
8.7 數(shù)據(jù)驅(qū)動方法················································································.176
8.7.1 方法一：原子空間中的聚類·····························································.176
8.7.2 方法二：邊緣檢測·······································································.180
8.8 計算重要性提議概率·······································································.180
8.9 計算多個不同的解··········································································.183
8.9.1 動機(jī)和數(shù)學(xué)原理·········································································.183
8.9.2 用于多種解的K-冒險家算法····························································.184
8.10 圖像分割實驗···············································································.185
8.11 應(yīng)用：圖像解析············································································.188
8.11.1 自上而下和自下而上的處理···························································.190
8.11.2 生成和判別方法········································································.190
8.11.3 馬爾可夫鏈核和子核···································································.191
8.11.4 DDMCMC 和提議概率·································································.193
8.11.5 馬爾可夫鏈子核········································································.200
8.11.6 圖像解析實驗···········································································.207
8.12 本章練習(xí)·····················································································.210
蒙特卡羅方法與人工智能
·X IV·
本章參考文獻(xiàn)·······················································································.211
第9 章 哈密頓和朗之萬蒙特卡羅方法····························································.215
9.1 引言····························································································.215
9.2 哈密頓力學(xué)···················································································.215
9.2.1 哈密頓方程···············································································.215
9.2.2 HMC 的簡單模型········································································.216
9.3 哈密頓力學(xué)的性質(zhì)··········································································.217
9.3.1 能量守恒·················································································.217
9.3.2 可逆性····················································································.218
9.3.3 辛結(jié)構(gòu)和體積保持·······································································.219
9.4 哈密頓方程的蛙跳離散化·································································.220
9.4.1 歐拉方法·················································································.220
9.4.2 改良的歐拉方法·········································································.220
9.4.3 蛙跳積分器···············································································.221
9.4.4 蛙跳積分器的特性·······································································.222
9.5 哈密頓蒙特卡羅方法和朗之萬蒙特卡羅方法·········································.223
9.5.1 HMC 建�！ぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁぁ�.223
9.5.2 HMC 算法················································································.224
9.5.3 LMC 算法················································································.226
9.5.4 HMC 調(diào)參················································································.228
9.5.5 HMC 的細(xì)致平衡證明···································································.229
9.6 黎曼流形HMC···············································································.230
9.6.1 HMC 中的線性變換·····································································.230
9.6.2 RMHMC 動力學(xué)·········································································.233
9.6.3 RMHMC 算法和變體····································································.235
9.6.4 RMHMC 中的協(xié)方差函數(shù)·······························································.236
9.7 HMC 實踐·····················································································.237
9.7.1 受約束正態(tài)分布的模擬實驗·····························································.237
9.7.2 使用RMHMC 對邏輯回歸系數(shù)進(jìn)行采樣···············································.241
9.7.3 使用LMC 采樣圖像密度：FRAME、GRADE 和DeepFRAME ·······················.243
9.8 本章練習(xí)······················································································.248
本章參考文獻(xiàn)·······················································································.249
第10 章 隨機(jī)梯度學(xué)習(xí)················································································.250
10.1 引言···························································································.250
目 錄
·XV·
10.2 隨機(jī)梯度：動機(jī)和性質(zhì)···································································.250
10.2.1 引例·····················································································.251
10.2.2 Robbins-Monro 定理····································································.253
10.2.3 隨機(jī)梯度下降和朗之萬方程···························································.254
10.3 馬爾可夫隨機(jī)場（MRF）模型的參數(shù)估計···········································.257
10.3.1 利用隨機(jī)梯度學(xué)習(xí)FRAME 模型······················································.258
10.3.2 FRAME 的替代學(xué)習(xí)方法·······························································.259
10.3.3 FRAME 算法的四種變體·······························································.261
10.3.4 紋理分析實驗···········································································.264
10.4 用神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)圖像模型································································.267
10.4.1 對比發(fā)散與持續(xù)對比發(fā)散······························································.267
10.4.2 使用深度網(wǎng)絡(luò)學(xué)習(xí)圖像的勢能模型：DeepFRAME···································.268
10.4.3 生成器網(wǎng)絡(luò)和交替反向傳播···························································.271
10.4.4 協(xié)作網(wǎng)絡(luò)和生成器模型································································.275
10.5 本章練習(xí)·····················································································.279
本章參考文獻(xiàn)·······················································································.279
第11 章 可視化能級圖················································································.282
11.1 引言···························································································.282
11.2 能級圖的示例、結(jié)構(gòu)和任務(wù)·····························································.282
11.2.1 基于能量的狀態(tài)空間劃分······························································.285
11.2.2 構(gòu)造非連通圖（DG）··································································.286
11.2.3 二維ELM 示例·········································································.287
11.2.4 表征學(xué)習(xí)任務(wù)的難度（或復(fù)雜度）····················································.289
11.3 廣義Wang-Landau 算法··································································.290
11.3.1 GWL 映射的能壘估計··································································.291
11.3.2 用GWL 估算體積······································································.292
11.3.3 GWL 收斂性分析·······································································.294
11.4 GWL 實驗···················································································.295
11.4.1 高斯混合模型的GWL 映射····························································.295
11.4.2 語法模型的GWL 映射·································································.301
11.5 用吸引-擴(kuò)散可視化能級圖······························································.305
11.5.1 亞穩(wěn)定性和宏觀劃分···································································.306
11.5.2 吸引-擴(kuò)散簡介·········································································.307
11.5.3 吸引-擴(kuò)散和Ising 模型································································.309
11.5.4 吸引-擴(kuò)散ELM 算法（ADELM 算法）···············································.311
蒙特卡羅方法與人工智能
·X VI·
11.5.5 調(diào)優(yōu)ADELM ···········································································.313
11.5.6 AD 能壘估計···········································································.314
11.6 用GWL 和ADELM 可視化SK 自旋玻璃模型······································.315
11.7 使用吸引?擴(kuò)散可視化圖像空間························································.318
11.7.1 圖像星系的結(jié)構(gòu)········································································.318
11.7.2 可視化實驗·············································································.319
11.8 本章練習(xí)·····················································································.324
本章參考文獻(xiàn)·······················································································.324