-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathrl-classification.html
1493 lines (918 loc) · 48.9 KB
/
rl-classification.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html class="theme-next mist use-motion" lang="zh-Hans">
<head><meta name="generator" content="Hexo 3.8.0">
<meta name="google-site-verification" content="zu-9nWphPjrzXV8v514mkHknIz4dNfHlib56-KNAu44">
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1">
<meta name="theme-color" content="#222">
<script src="/lib/pace/pace.min.js?v=1.0.2"></script>
<link href="/lib/pace/pace-theme-flash.min.css?v=1.0.2" rel="stylesheet">
<meta http-equiv="Cache-Control" content="no-transform">
<meta http-equiv="Cache-Control" content="no-siteapp">
<script>
(function(i,s,o,g,r,a,m){i["DaoVoiceObject"]=r;i[r]=i[r]||function(){(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;a.charset="utf-8";m.parentNode.insertBefore(a,m)})(window,document,"script",('https:' == document.location.protocol ? 'https:' : 'http:') + "//widget.daovoice.io/widget/356f1943.js","daovoice")
daovoice('init', {
app_id: "356f1943"
});
daovoice('update');
</script>
<link href="/lib/fancybox/source/jquery.fancybox.css?v=2.1.5" rel="stylesheet" type="text/css">
<link href="/lib/font-awesome/css/font-awesome.min.css?v=4.6.2" rel="stylesheet" type="text/css">
<link href="/css/main.css?v=5.1.4" rel="stylesheet" type="text/css">
<link rel="apple-touch-icon" sizes="180x180" href="/images/apple-touch-icon.png?v=5.1.4">
<link rel="icon" type="image/png" sizes="32x32" href="/images/favicon-32x32png?v=5.1.4">
<link rel="icon" type="image/png" sizes="16x16" href="/images/favicon-16x16.png?v=5.1.4">
<link rel="mask-icon" href="/images/logo.svg?v=5.1.4" color="#222">
<meta name="keywords" content="rl,">
<link rel="alternate" href="/atom.xml" title="Keavnn'Blog" type="application/atom+xml">
<script>
(function(){
if(''){
if (prompt('请输入文章密码','') !== ''){
alert('密码错误！');
history.back();
}
}
})();
</script>
<meta name="description" content="本文讲述了强化学习中各种算法、问题的分类规则。">
<meta name="keywords" content="rl">
<meta property="og:type" content="article">
<meta property="og:title" content="强化学习的类别">
<meta property="og:url" content="http://StepNeverStop.github.io/rl-classification.html">
<meta property="og:site_name" content="Keavnn'Blog">
<meta property="og:description" content="本文讲述了强化学习中各种算法、问题的分类规则。">
<meta property="og:locale" content="zh-Hans">
<meta property="og:image" content="http://stepneverstop.github.io/rl-classification/stationary.png">
<meta property="og:image" content="http://stepneverstop.github.io/rl-classification/non-stationary.png">
<meta property="og:image" content="http://stepneverstop.github.io/rl-classification/model-classification.png">
<meta property="og:image" content="http://stepneverstop.github.io/rl-classification/policy-based.png">
<meta property="og:image" content="http://stepneverstop.github.io/rl-classification/value-based.png">
<meta property="og:updated_time" content="2019-09-03T10:27:17.253Z">
<meta name="twitter:card" content="summary">
<meta name="twitter:title" content="强化学习的类别">
<meta name="twitter:description" content="本文讲述了强化学习中各种算法、问题的分类规则。">
<meta name="twitter:image" content="http://stepneverstop.github.io/rl-classification/stationary.png">
<script type="text/javascript" id="hexo.configurations">
var NexT = window.NexT || {};
var CONFIG = {
root: '/',
scheme: 'Mist',
version: '5.1.4',
sidebar: {"position":"left","display":"post","offset":12,"b2t":false,"scrollpercent":true,"onmobile":true},
fancybox: true,
tabs: true,
motion: {"enable":true,"async":true,"transition":{"post_block":"fadeIn","post_header":"slideDownIn","post_body":"slideDownIn","coll_header":"slideLeftIn","sidebar":"slideUpIn"}},
duoshuo: {
userId: '0',
author: '博主'
},
algolia: {
applicationID: '',
apiKey: '',
indexName: '',
hits: {"per_page":10},
labels: {"input_placeholder":"Search for Posts","hits_empty":"We didn't find any results for the search: ${query}","hits_stats":"${hits} results found in ${time} ms"}
}
};
</script>
<link rel="canonical" href="http://StepNeverStop.github.io/rl-classification.html">
<title>强化学习的类别 | Keavnn'Blog</title>
</head>
<body itemscope="" itemtype="http://schema.org/WebPage" lang="zh-Hans">
<div class="container sidebar-position-left page-post-detail">
<div class="headband"></div>
<a href="https://github.com/StepNeverStop" class="github-corner" aria-label="View source on GitHub" rel="external nofollow" target="_blank"><svg width="80" height="80" viewbox="0 0 250 250" style="fill:#151513; color:#fff; position: absolute; top: 0; border: 0; right: 0;" aria-hidden="true"><path d="M0,0 L115,115 L130,115 L142,142 L250,250 L250,0 Z"/><path d="M128.3,109.0 C113.8,99.7 119.0,89.6 119.0,89.6 C122.0,82.7 120.5,78.6 120.5,78.6 C119.2,72.0 123.4,76.3 123.4,76.3 C127.3,80.9 125.5,87.3 125.5,87.3 C122.9,97.6 130.6,101.9 134.4,103.2" fill="currentColor" style="transform-origin: 130px 106px;" class="octo-arm"/><path d="M115.0,115.0 C114.9,115.1 118.7,116.5 119.8,115.4 L133.7,101.6 C136.9,99.2 139.9,98.4 142.2,98.6 C133.8,88.0 127.5,74.4 143.8,58.0 C148.5,53.4 154.0,51.2 159.7,51.0 C160.3,49.4 163.2,43.6 171.4,40.1 C171.4,40.1 176.1,42.5 178.8,56.2 C183.1,58.6 187.2,61.8 190.9,65.4 C194.5,69.0 197.7,73.2 200.1,77.6 C213.8,80.2 216.3,84.9 216.3,84.9 C212.7,93.1 206.9,96.0 205.4,96.6 C205.1,102.4 203.0,107.8 198.3,112.5 C181.9,128.9 168.3,122.5 157.7,114.1 C157.9,116.9 156.7,120.9 152.7,124.9 L141.0,136.5 C139.8,137.7 141.6,141.9 141.8,141.8 Z" fill="currentColor" class="octo-body"/></svg></a><style>.github-corner:hover .octo-arm{animation:octocat-wave 560ms ease-in-out}@keyframes octocat-wave{0%,100%{transform:rotate(0)}20%,60%{transform:rotate(-25deg)}40%,80%{transform:rotate(10deg)}}@media (max-width:500px){.github-corner:hover .octo-arm{animation:none}.github-corner .octo-arm{animation:octocat-wave 560ms ease-in-out}}</style>
<header id="header" class="header" itemscope="" itemtype="http://schema.org/WPHeader">
<div class="header-inner"><div class="site-brand-wrapper">
<div class="site-meta ">
<div class="custom-logo-site-title">
<a href="/" class="brand" rel="start">
<span class="logo-line-before"><i></i></span>
<span class="site-title">Keavnn'Blog</span>
<span class="logo-line-after"><i></i></span>
</a>
</div>
<h1 class="site-subtitle" itemprop="description">If it is to be, it is up to me.</h1>
</div>
<div class="site-nav-toggle">
<button>
<span class="btn-bar"></span>
<span class="btn-bar"></span>
<span class="btn-bar"></span>
</button>
</div>
</div>
<nav class="site-nav">
<ul id="menu" class="menu">
<li class="menu-item menu-item-home">
<a href="/" rel="section">
<i class="menu-item-icon fa fa-fw fa-home"></i> <br>
首页
</a>
</li>
<li class="menu-item menu-item-about">
<a href="/about/" rel="section">
<i class="menu-item-icon fa fa-fw fa-user"></i> <br>
关于
</a>
</li>
<li class="menu-item menu-item-tags">
<a href="/tags/" rel="section">
<i class="menu-item-icon fa fa-fw fa-tags"></i> <br>
标签
</a>
</li>
<li class="menu-item menu-item-categories">
<a href="/categories/" rel="section">
<i class="menu-item-icon fa fa-fw fa-th"></i> <br>
分类
</a>
</li>
<li class="menu-item menu-item-archives">
<a href="/archives/" rel="section">
<i class="menu-item-icon fa fa-fw fa-archive"></i> <br>
归档
</a>
</li>
<li class="menu-item menu-item-search">
<a href="javascript:;" class="popup-trigger">
<i class="menu-item-icon fa fa-search fa-fw"></i> <br>
搜索
</a>
</li>
</ul>
<div class="site-search">
<div class="popup search-popup local-search-popup">
<div class="local-search-header clearfix">
<span class="search-icon">
<i class="fa fa-search"></i>
</span>
<span class="popup-btn-close">
<i class="fa fa-times-circle"></i>
</span>
<div class="local-search-input-wrapper">
<input autocomplete="off" placeholder="搜索..." spellcheck="false" type="text" id="local-search-input">
</div>
</div>
<div id="local-search-result"></div>
</div>
</div>
</nav>
</div>
</header>
<main id="main" class="main">
<div class="main-inner">
<div class="content-wrap">
<div id="content" class="content">
<div id="posts" class="posts-expand">
<article class="post post-type-normal" itemscope="" itemtype="http://schema.org/Article">
<div class="post-block">
<link itemprop="mainEntityOfPage" href="http://StepNeverStop.github.io/rl-classification.html">
<span hidden itemprop="author" itemscope="" itemtype="http://schema.org/Person">
<meta itemprop="name" content="Keavnn">
<meta itemprop="description" content="">
<meta itemprop="image" content="/images/Kicon.jpg">
</span>
<span hidden itemprop="publisher" itemscope="" itemtype="http://schema.org/Organization">
<meta itemprop="name" content="Keavnn'Blog">
</span>
<header class="post-header">
<h2 class="post-title" itemprop="name headline">强化学习的类别</h2>
<div class="post-meta">
<span class="post-time">
<span class="post-meta-item-icon">
<i class="fa fa-calendar-o"></i>
</span>
<span class="post-meta-item-text">发表于</span>
<time title="创建于" itemprop="dateCreated datePublished" datetime="2019-05-10T15:18:23+08:00">
2019-05-10
</time>
<span class="post-meta-divider">|</span>
<span class="post-meta-item-icon">
<i class="fa fa-calendar-check-o"></i>
</span>
<span class="post-meta-item-text">更新于:</span>
<time title="更新于" itemprop="dateModified" datetime="2019-09-03T18:27:17+08:00">
2019-09-03
</time>
</span>
<span class="post-category">
<span class="post-meta-divider">|</span>
<span class="post-meta-item-icon">
<i class="fa fa-folder-o"></i>
</span>
<span class="post-meta-item-text">分类于</span>
<span itemprop="about" itemscope="" itemtype="http://schema.org/Thing">
<a href="/categories/ReinforcementLearning/" itemprop="url" rel="index">
<span itemprop="name">ReinforcementLearning</span>
</a>
</span>
</span>
<div class="post-wordcount">
<span class="post-meta-item-icon">
<i class="fa fa-file-word-o"></i>
</span>
<span class="post-meta-item-text">字数统计:</span>
<span title="字数统计">
1.8k
</span>
<span class="post-meta-divider">|</span>
<span class="post-meta-item-icon">
<i class="fa fa-clock-o"></i>
</span>
<span class="post-meta-item-text">阅读时长 ≈</span>
<span title="阅读时长">
7
</span>
</div>
</div>
</header>
<div class="post-body" itemprop="articleBody">
<p>本文讲述了强化学习中各种算法、问题的分类规则。</p>
<a id="more"></a>
<h1 id="Stationary-or-not"><a href="#Stationary-or-not" class="headerlink" title="Stationary or not"></a>Stationary or not</h1><p>根据环境十分稳定、可以将强化学习问题分为stationary、non-stationary。</p>
<p>如果状态转移<strong>和</strong>奖励函数是确定的，即选择动作$a$后执行它的结果是确定的，那么这个环境就是stationary。</p>
<p><img src="./rl-classification/stationary.png" alt=""></p>
<p>如果状态转移<strong>或</strong>奖励函数是不确定的，即选择动作$a$后执行它的结果是不确定的，那么这个环境就是non-stationary。</p>
<p><img src="./rl-classification/non-stationary.png" alt=""></p>
<h1 id="Model-Based-or-Free"><a href="#Model-Based-or-Free" class="headerlink" title="Model Based-or-Free"></a>Model Based-or-Free</h1><p>一直对这个问题的认识不清晰，直到最近（2019年5月12日19:13:53）才有了清晰的认识。</p>
<p>需要注意的是，无论是Model-Based还是Model-Free都不是对强化学习问题的分类，而是对算法的分类。之前一直理解的是状态空间$\mathcal{S}$、动作空间$\mathcal{A}$的都是离散的，转移概率矩阵$\mathcal{P}$是确定的，这样即是Model-Based，如果状态空间$\mathcal{S}$、动作空间$\mathcal{A}$或转移概率矩阵$\mathcal{P}$是不确定的，则是Model-Free，其实这只是对Model的分类，并不是Model-Based与Model-Free的真实含义，Model-Based与Model-Free是对算法求解过程的分类，理解这个可以在阅读国外文献、实验环境时更清晰，提升自己对强化学习算法的理解深度。</p>
<p>Model-Based：</p>
<ul>
<li>智能体Agent在已知模型（$\mathcal{S,A,R,P}$有限且确定）或者先学习一个模型（使用有监督对状态转移、奖励函数进行学习而得到），并在这个模型中使用<strong>planning</strong>（预测所有状态转移可能）方法来计算解决方案</li>
</ul>
<blockquote>
<p>Now if we know what all those elements of an MDP are, we can just compute the solution before ever actually executing an action in the environment. In AI, we typically call computing the solution to a decision-making problem before executing an actual decision <em>planning</em>. Some classic planning algorithms for MDPs include Value Iteration, Policy Iteration, and whole lot more.</p>
</blockquote>
<p>Model-Free:</p>
<ul>
<li>智能体在模型（$\mathcal{S,A,R,P}$可能确定但没有使用planning方式解决，也可能不确定）中试错，并且使用<strong>learning</strong>（不预测全部可能性）方法来产生最佳策略</li>
</ul>
<blockquote>
<p>But the RL problem isn’t so kind to us. What makes a problem an RL problem, rather than a planning problem, is the agent does <em>not</em> know all the elements of the MDP, precluding it from being able to plan a solution. Specifically, the agent does not know how the world will change in response to its actions (the transition function TT), nor what immediate reward it will receive for doing so (the reward function RR). The agent will simply have to try taking actions in the environment, observe what happens, and somehow, find a good policy from doing so.</p>
</blockquote>
<p>根据Model-Based、Model-Free对算法、解决方法进行分类：</p>
<p>Model-Based：DP、Policy Iteration、Value Iteration……</p>
<p>Model-Free：SARSA、Q-Learning、PG……</p>
<hr>
<blockquote>
<p>if you want a way to check if an RL algorithm is model-based or model-free, ask yourself this question: after learning, can the agent make predictions about what the next state and reward will be before it takes each action? If it can, then it’s a model-based RL algorithm. if it cannot, it’s a model-free algorithm.</p>
</blockquote>
<p><strong>使用算法学成策略之后，智能体可以在执行动作前判断该动作的后果，即是Model-Based，反之则是Model-Free</strong></p>
<hr>
<p>根据Model-Based和Model-Free可以将强化学习算法分类，图片摘自OpenAI Spinning Up，如图所示：</p>
<p><img src="./rl-classification/model-classification.png" alt=""></p>
<blockquote>
<p><a href="https://www.quora.com/What-is-the-difference-between-model-based-and-model-free-reinforcement-learning" rel="external nofollow" target="_blank">What is the difference between model-based and model-free reinforcement learning?</a></p>
<p><a href="https://spinningup.openai.com/en/latest/spinningup/rl_intro2.html#a-taxonomy-of-rl-algorithms" rel="external nofollow" target="_blank">OpenAI Spinning Up : A Taxonomy of RL Algorithms</a></p>
</blockquote>
<h1 id="Policy-or-Value"><a href="#Policy-or-Value" class="headerlink" title="Policy or Value"></a>Policy or Value</h1><p>强化学习的目的是找到最优策略使得累积期望回报最大化，获得最优策略的方法有直接与间接之分。直接获取策略的图示为：</p>
<p><img src="./rl-classification/policy-based.png" alt=""></p>
<p>间接获得策略为从值函数中提取最优策略，图示为：</p>
<p><img src="./rl-classification/value-based.png" alt=""></p>
<p>直接获取策略的方式即为Policy-Based，常见的算法有：</p>
<ul>
<li>Policy Gradient</li>
<li>PPO</li>
<li>SAC</li>
<li>……</li>
</ul>
<p>间接获得策略的方式即为Value-Based，常见的算法有：</p>
<ul>
<li>SARSA</li>
<li>Q-Learning</li>
<li>DQN</li>
<li>……</li>
</ul>
<h1 id="On-policy-or-Off-policy"><a href="#On-policy-or-Off-policy" class="headerlink" title="On-policy or Off-policy"></a>On-policy or Off-policy</h1><p>在机器学习中，提到On跟Off这两个词我们最容易想到的是On-line Learning与Off-line Learning，那么强化学习与On-line、Off-line有什么关系呢？</p>
<hr>
<p>网上对于On-line Learning与Off-line Learning有不同的解释，按热度排序为下面三种：</p>
<p>对于On-line Learning：</p>
<ol>
<li>单样本学习，样本用完即丢，样本连续不断输入，非数据集，而是数据流</li>
<li>单样本的（SGD）</li>
<li>单样本或批样本学习，样本连续不断输入，非数据集，而是数据流</li>
</ol>
<p>相应对于Off-line Learning：</p>
<ol>
<li>批样本或全样本学习多次，静态样本集</li>
<li>批样本学习</li>
<li>全样本学习，静态数据集</li>
</ol>
<p>对于这三种方式，强化学习可以怎样融入呢？</p>
<ol>
<li>对于第一种，强化学习不属于On-line Learning也不属于Off-line Learning，不属于Off-line Learning是因为样本非静态、非固定，不属于On-line Learning是因为对于Q-Learning、Sarsa、PG、PPO等算法样本用完即丢，对于DQN、TD3等算法样本重复利用。</li>
<li>对于第二种，强化学习包括On-line Learning及On-line Learning</li>
<li>低于第三种，强化学习属于On-line Learning</li>
</ol>
<hr>
<p>On-policy、Off-policy与On-line、Off-line之间有关系吗？</p>
<p>好像没有关系。虽然它们都是关于样本进行的划分，不过On-Off line learning针对的是样本的使用，而On-Off policy针对的是样本的生成。</p>
<hr>
<p>学习On-policy、Off-policy之前首先需要理解什么是行为策略与目标策略。</p>
<p>行为策略$\mathcal{Behavior Policy}$：</p>
<ul>
<li>采样时间序列$S_{0},A_{0},R_{0},S_{1},A_{1},R_{2},…,S_{n},A_{n},R_{n}$的策略</li>
<li>官话：指导个体产生与环境进行实际交互行为的策略</li>
<li>未必由一个模型表示</li>
</ul>
<p>目标策略$\mathcal{TargetPolicy}$:</p>
<ul>
<li>待优化的策略</li>
<li>官话：用来评价状态或行为价值的策略或者待优化的策略称为目标策略</li>
</ul>
<p>同步策略学习$\mathcal{On-Policy}$:</p>
<ul>
<li>简言之，边采样边学习</li>
<li>官话：如果个体在学习过程中优化的策略与自己的行为策略是同一个策略时，这种学习方式称为<strong>同步策略学习（on-policy learning）</strong></li>
<li>行为策略与目标策略是同一个</li>
</ul>
<p>异步策略学习$\mathcal{Off-Policy}$:</p>
<ul>
<li>简言之，你采样我学习</li>
<li>官话：如果个体在学习过程中优化的策略与自己的行为策略是不同的策略时，这种学习方式称为<strong>异步策略学习（off-policy learning）</strong></li>
<li>行为策略与目标策略不同，行为策略可能是目标策略的“分身”（双网络结构），或者完全是另一个采样的策略</li>
</ul>
<p>两者的区别简而言之：如果需要估计一个值，用于估计的额外信息和当前信息出自同一策略则为on-policy，否则为off-policy。以SARSA和Q-Learning算法为例，对于Q值的估计，SARSA中$s_{t+1}$的动作由当前策略产生，故为on-policy算法，而Q-Learning中$s_{t+1}$的动作由贪心策略产生，故为off-policy。</p>
<p>例如：</p>
<div class="table-container">
<table>
<thead>
<tr>
<th style="text-align:center"></th>
<th style="text-align:center">SARSA</th>
<th style="text-align:center">Q-learning</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align:center">Choosing A’</td>
<td style="text-align:center">π</td>
<td style="text-align:center">π</td>
</tr>
<tr>
<td style="text-align:center">Updating Q</td>
<td style="text-align:center">π</td>
<td style="text-align:center">μ</td>
</tr>
</tbody>
</table>
</div>
<blockquote>
<p><a href="https://stackoverflow.com/a/41420616" rel="external nofollow" target="_blank">一个以Q-Learning和Sarsa算法做比较的解释</a></p>
</blockquote>
<h1 id="Stochastic-or-Deterministic"><a href="#Stochastic-or-Deterministic" class="headerlink" title="Stochastic or Deterministic"></a>Stochastic or Deterministic</h1>
</div>
<div>
<div>
<div style="text-align:center;color: #ccc;font-size:14px;">-------------本文结束<i class="fa fa-heart"></i>感谢您的阅读-------------</div>
</div>
</div>
<div>
<div class="my_post_copyright">
<script src="//cdn.bootcss.com/clipboard.js/1.5.10/clipboard.min.js"></script>
<!-- JS库 sweetalert 可修改路径 -->
<script src="https://cdn.bootcss.com/jquery/2.0.0/jquery.min.js"></script>
<script src="https://unpkg.com/sweetalert/dist/sweetalert.min.js"></script>
<p><span>本文标题:</span><a href="/rl-classification.html">强化学习的类别</a></p>
<p><span>文章作者:</span><a href="/" title="访问 Keavnn 的个人博客">Keavnn</a></p>
<p><span>发布时间:</span>2019年05月10日 - 15:05</p>
<p><span>最后更新:</span>2019年09月03日 - 18:09</p>
<p><span>原始链接:</span><a href="/rl-classification.html" title="强化学习的类别">http://StepNeverStop.github.io/rl-classification.html</a>
<span class="copy-path" title="点击复制文章链接"><i class="fa fa-clipboard" data-clipboard-text="http://StepNeverStop.github.io/rl-classification.html" aria-label="复制成功！"></i></span>
</p>
<p><span>许可协议:</span><i class="fa fa-creative-commons"></i> <a rel="external nofollow" href="https://creativecommons.org/licenses/by-nc-sa/4.0/" target="_blank" title="Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)">署名-非商业性使用-相同方式共享 4.0 国际</a> 转载请保留原文链接及作者。</p>
</div>
<script>
var clipboard = new Clipboard('.fa-clipboard');
$(".fa-clipboard").click(function(){
clipboard.on('success', function(){
swal({
title: "",
text: '复制成功',
icon: "success",
showConfirmButton: true
});
});
});
</script>
</div>
<div>
<div style="padding: 10px 0; margin: 20px auto; width: 90%; text-align: center;">
<div>如果您获得了帮助，也可以资助一下小的啦~</div>
<button id="rewardButton" disable="enable" onclick="var qr = document.getElementById('QR'); if (qr.style.display === 'none') {qr.style.display='block';} else {qr.style.display='none'}">
<span>打赏啦</span>
</button>
<div id="QR" style="display: none;">
<div id="wechat" style="display: inline-block">
<img id="wechat_qr" src="/images/wechatpay.jpg" alt="Keavnn 微信">
<p>微信</p>
</div>
<div id="alipay" style="display: inline-block">
<img id="alipay_qr" src="/images/alipay.jpg" alt="Keavnn 支付宝">
<p>支付宝</p>
</div>
</div>
</div>
</div>
<footer class="post-footer">
<div class="post-tags">
<a href="/tags/rl/" rel="tag"> <i class="fa fa-tag"></i> rl</a>
</div>
<div class="post-nav">
<div class="post-nav-next post-nav-item">
<a href="/价值与贝尔曼方程.html" rel="next" title="价值与贝尔曼方程">
<i class="fa fa-chevron-left"></i> 价值与贝尔曼方程
</a>
</div>
<span class="post-nav-divider"></span>
<div class="post-nav-prev post-nav-item">
<a href="/dynamic-programming.html" rel="prev" title="动态规划 Dynamic Programming">
动态规划 Dynamic Programming <i class="fa fa-chevron-right"></i>
</a>
</div>
</div>
</footer>
</div>
</article>
<div class="post-spread">
<!-- Go to www.addthis.com/dashboard to customize your tools -->
<div class="addthis_inline_share_toolbox">
<script type="text/javascript" src="//s7.addthis.com/js/300/addthis_widget.js#pubid=ra-5cefbfc88c13b0e7" async="async"></script>
</div>
</div>
</div>
</div>
<div class="comments" id="comments">
<div id="lv-container" data-id="city" data-uid="MTAyMC80MTk0NS8xODQ5MQ=="></div>
</div>
</div>
<div class="sidebar-toggle">
<div class="sidebar-toggle-line-wrap">
<span class="sidebar-toggle-line sidebar-toggle-line-first"></span>
<span class="sidebar-toggle-line sidebar-toggle-line-middle"></span>
<span class="sidebar-toggle-line sidebar-toggle-line-last"></span>
</div>
</div>
<aside id="sidebar" class="sidebar">
<div id="sidebar-dimmer"></div>
<div class="sidebar-inner">
<ul class="sidebar-nav motion-element">
<li class="sidebar-nav-toc sidebar-nav-active" data-target="post-toc-wrap">
文章目录
</li>
<li class="sidebar-nav-overview" data-target="site-overview-wrap">
站点概览
</li>
</ul>
<section class="site-overview-wrap sidebar-panel">
<div class="site-overview">
<div class="site-author motion-element" itemprop="author" itemscope="" itemtype="http://schema.org/Person">
<img class="site-author-image" itemprop="image" src="/images/Kicon.jpg" alt="Keavnn">
<p class="site-author-name" itemprop="name">Keavnn</p>
<p class="site-description motion-element" itemprop="description">If it is to be, it is up to me.</p>
</div>
<nav class="site-state motion-element">
<div class="site-state-item site-state-posts">
<a href="/archives/">
<span class="site-state-item-count">51</span>
<span class="site-state-item-name">日志</span>
</a>
</div>
<div class="site-state-item site-state-categories">
<a href="/categories/index.html">
<span class="site-state-item-count">11</span>
<span class="site-state-item-name">分类</span>
</a>
</div>
<div class="site-state-item site-state-tags">
<a href="/tags/index.html">
<span class="site-state-item-count">26</span>
<span class="site-state-item-name">标签</span>
</a>
</div>
</nav>
<div class="feed-link motion-element">
<a href="/atom.xml" rel="alternate">
<i class="fa fa-rss"></i>
RSS
</a>
</div>
<div class="links-of-author motion-element">
<span class="links-of-author-item">
<a href="https://github.com/StepNeverStop" target="_blank" title="GitHub" rel="external nofollow">
<i class="fa fa-fw fa-github"></i>GitHub</a>
</span>
<span class="links-of-author-item">
<a href="mailto:[email protected]" target="_blank" title="E-Mail" rel="external nofollow">
<i class="fa fa-fw fa-envelope"></i>E-Mail</a>
</span>
</div>
<div class="cc-license motion-element" itemprop="license">
<a href="https://creativecommons.org/licenses/by-nc-sa/4.0/" class="cc-opacity" target="_blank" rel="external nofollow">
<img src="/images/cc-by-nc-sa.svg" alt="Creative Commons">
</a>
</div>
<div class="links-of-blogroll motion-element links-of-blogroll-inline">
<div class="links-of-blogroll-title">
<i class="fa fa-fw fa-link"></i>
推荐阅读
</div>
<ul class="links-of-blogroll-list">
<li class="links-of-blogroll-item">
<a href="https://bluefisher.github.io" title="Fisher Chang" target="_blank" rel="external nofollow">Fisher Chang</a>
</li>
</ul>
</div>
</div>
</section>
<!--noindex-->
<section class="post-toc-wrap motion-element sidebar-panel sidebar-panel-active">
<div class="post-toc">
<div class="post-toc-content"><ol class="nav"><li class="nav-item nav-level-1"><a class="nav-link" href="#Stationary-or-not"><span class="nav-number">1.</span> <span class="nav-text">Stationary or not</span></a></li><li class="nav-item nav-level-1"><a class="nav-link" href="#Model-Based-or-Free"><span class="nav-number">2.</span> <span class="nav-text">Model Based-or-Free</span></a></li><li class="nav-item nav-level-1"><a class="nav-link" href="#Policy-or-Value"><span class="nav-number">3.</span> <span class="nav-text">Policy or Value</span></a></li><li class="nav-item nav-level-1"><a class="nav-link" href="#On-policy-or-Off-policy"><span class="nav-number">4.</span> <span class="nav-text">On-policy or Off-policy</span></a></li><li class="nav-item nav-level-1"><a class="nav-link" href="#Stochastic-or-Deterministic"><span class="nav-number">5.</span> <span class="nav-text">Stochastic or Deterministic</span></a></li></ol></div>
</div>
</section>
<!--/noindex-->
</div>
</aside>
</div>
</main>
<footer id="footer" class="footer">
<div class="footer-inner">
<script async src="https://busuanzi.ibruce.info/busuanzi/2.3/busuanzi.pure.mini.js"></script>
<div class="copyright">© <span itemprop="copyrightYear">2020</span>
<span class="with-love">
<i class="fa fa-heart"></i>
</span>
<span class="author" itemprop="copyrightHolder">Keavnn</span>
<span class="post-meta-divider">|</span>
<span class="post-meta-item-icon">
<i class="fa fa-area-chart"></i>
</span>
<span class="post-meta-item-text">Site words total count:</span>
<span title="Site words total count">80.3k</span>
</div>
<div class="powered-by">
<i class="fa fa-user-md"></i><span id="busuanzi_container_site_pv">
本站总访问量<span id="busuanzi_value_site_pv"></span>次
</span>
</div>
<!-- <div class="theme-info">
<div class="powered-by"></div>
<span class="post-count">博客全站共80.3k字</span>
</div> -->
</div>
</footer>
<div class="back-to-top">
<i class="fa fa-arrow-up"></i>
<span id="scrollpercent"><span>0</span>%</span>
</div>
</div>
<script type="text/javascript">
if (Object.prototype.toString.call(window.Promise) !== '[object Function]') {
window.Promise = null;
}
</script>