-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
371 lines (334 loc) · 15.6 KB
/
index.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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description"
content="We proposed closed-form policy improvement operators and modeled the behavior policies as a Gaussian Mixture.">
<meta name="keywords" content=" Offline Reinforcement Learning algorithms,Deep Reinforcement Learning">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Offline Reinforcement Learning with Closed-Form Policy Improvement Operators
</title>
<!-- custom fonts -->
<link rel="stylesheet" type="text/css"
href="https://cdn.rawgit.com/dreampulse/computer-modern-web-font/master/fonts.css">
<link href="https://fonts.cdnfonts.com/css/proxima-nova-2" rel="stylesheet">
<!-- end custom fonts -->
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="icon" type="ico" href="./static/favicon.ico">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./static/js/index.js"></script>
</head>
<body>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<div class="navbar-item has-dropdown is-hoverable">
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<a class="navbar-item" href="https://sites.google.com/eng.ucsd.edu/multi-task-batch-reinforcement/home">
Metric Learning Offline RL
</a>
<a class="navbar-item" href="https://arxiv.org/abs/2110.12081.pdf">
Off-policy RL with Optimism and DICE
</a>
<a class="navbar-item" href="https://lcd.eddie.win">
Language Control Diffusion
</a>
</div>
</div>
</div>
</div>
</nav>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title">Offline Reinforcement Learning with Closed-Form Policy Improvement
Operators
</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://sites.google.com/view/jiachenli/home">Jiachen Li</a><sup>*</sup><sup>1</sup>,</span>
<span class="author-block">
<a href="https://eddie.win">Edwin Zhang</a><sup>*</sup><sup>1,3</sup>,</span>
<span class="author-block">
<a href="https://mingyin0312.github.io/">Ming Yin</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="https://www.cs.bu.edu/groups/ivc/qinxun/">Qinxun Bai</a><sup>2</sup>,
</span>
<span class="author-block">
<a href="https://sites.cs.ucsb.edu/~yuxiangw/">Yu-Xiang Wang</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="https://sites.cs.ucsb.edu/~william/">William Yang Wang</a><sup>1</sup>
</span>
</div>
<div class="is-size-5 publication-authors" style="margin-top: 15px;">
<span class="author-block"><sup>1</sup>UC Santa Barbara,</span>
<span class="author-block"><sup>2</sup>Horizon Robotics,</span>
<span class="author-block"><sup>3</sup>Harvard University</span>
</div>
<div class="publication-authors">
<span class="author-block"><sup>*</sup>Equal Contribution</span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a target="_blank" href="https://proceedings.mlr.press/v202/li23av/li23av.pdf" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a target="_blank" href="https://arxiv.org/abs/2211.15956"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Code Link. -->
<span class="link-block">
<a target="_blank" href="https://github.com/cfpi-icml23/cfpi"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<!-- Dataset Link. -->
<span class="link-block">
<a target="_blank" href="./static/images/poster.png"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="far fa-images"></i>
</span>
<span>Poster</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body">
<img src="./static/images/1.png" alt="" srcset="">
<h2 class="subtitle has-text-centered">
<span class="dnerf">Closed-form policy improvement operators</span> enable stable offline policy improvement.
</h2>
</div>
</div>
</section>
<section class="hero is-light is-small">
<div class="hero-body">
<div class="container is-max-desktop">
<div id="results-carousel" class="carousel results-carousel">
<div class="item">
<img src="./static/images/figures/iterative_pac/antmaze-large-diverse-v0.png" style="width: auto;" alt="">j
</div>
<div class="item">
<img src="./static/images/figures/iterative_pac/antmaze-large-play-v0.png" style="width: auto;" alt="">j
</div>
<div class="item">
<img src="./static/images/figures/iterative_pac/antmaze-medium-diverse-v0.png" style="width: auto;" alt="">
</div>
<div class="item">
<img src="./static/images/figures/iterative_pac/antmaze-medium-play-v0.png" style="width: auto;" alt="">
</div>
<div class="item">
<img src="./static/images/figures/iterative_pac/antmaze-umaze-diverse-v0.png" style="width: auto;" alt="">
</div>
<div class="item">
<img src="./static/images/figures/iterative_pac/antmaze-umaze-v0.png" style="width: auto;" alt="">
</div>
</div>
</div>
<div class="container is-max-desktop">
<img style="border: 1px solid black; padding: 10px; background: white; margin-top: 40px; border-radius: 10px;"
src="./static/images/2.png" width="100%" alt="">
</div>
<div class="has-text-centered" style="margin: 10px 0">
Our method's offline training results on AntMaze. Shaded area denotes one standard deviation.
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<!-- Abstract. -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Abstract</h2>
<div class="content has-text-justified">
<p>
Behavior constrained policy optimization has been demonstrated to be a successful paradigm for tackling
Offline
Reinforcement Learning. By exploiting historical transitions, a policy is trained to maximize a learned
value function
while constrained by the behavior policy to avoid a significant distributional shift. </p>
<p>
In this project, we
propose <span class="dnerf"> closed-form policy improvement operators</span>. We make the novel
observation
that the behavior constraint
naturally motivates
the use of first-order Taylor approximation, leading to a linear approximation of the policy objective.
</p>
<p>
Additionally, as
practical datasets are usually collected by heterogeneous policies, we model the behavior policies as a
Gaussian Mixture
and overcome the induced optimization difficulties by leveraging the LogSumExp's lower bound and Jensen's
Inequality,
giving rise to a closed-form policy improvement operator.</p>
<p>
We instantiate both one-step and iterative
offline RL
algorithms with our novel policy improvement operators and empirically demonstrate {their} effectiveness
over
state-of-the-art algorithms on the standard D4RL benchmark.
</p>
</div>
</div>
</div>
<!--/ Abstract. -->
<!-- Paper video. -->
<div class="columns is-centered has-text-centered">
<div class="column" style="height: 100%">
<h2 class="title is-3">Video</h2>
<div class="publication-video h750">
<iframe
src="https://slideslive.com/embed/presentation/38993793?auto_load=true&auto_play=false&zoom_ratio=&disable_fullscreen=false&locale=en&vertical_enabled=true&vertical_enabled_on_mobile=false&allow_hidden_controls_when_paused=true&fit_to_viewport=true&custom_user_id=&user_uuid=df1abce8-48c1-4383-828e-a2574e02127f"
scrolling="no" frameborder="0"
sandbox="allow-forms allow-pointer-lock allow-popups allow-same-origin allow-scripts allow-top-navigation"
allow="autoplay; fullscreen" allowfullscreen="" webkitallowfullscreen="" mozallowfullscreen=""
class="h750" style="margin: 0px auto; display: block; width: 100%;"></iframe>
<!-- <iframe src="https://www.youtube.com/embed/MrKrnHhk8IA?rel=0&showinfo=0" frameborder="0"
allow="autoplay; encrypted-media" allowfullscreen></iframe> -->
</div>
</div>
</div>
<!--/ Paper video. -->
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<!-- Concurrent Work. -->
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Statistical Evaluation</h2>
<img src="./static/images/figures/score_dist.png" alt="" srcset="">
<img src="./static/images/figures/reliable_eval.png" alt="" srcset="">
<div class="content has-text-justified">
<p>Here, I-MG (MG-PAC) and I-SG (SG-PAC) refer to our method.</p>
<p>
<a href="https://arxiv.org/abs/2108.13264" target="_blank">Performance profiles</a> (score distributions)
for all methods on the 9 tasks from the D4RL MuJoCo Gym
domain.
The average score is calculated by averaging all runs within one task. Each task contains 10 seeds, and
each seed
evaluates
for 100 episodes. Shaded area denotes 95% confidence bands based on percentile bootstrap and stratified
sampling. The η value where the curves intersect with the dashed horizontal line y = 0.5 corresponds
to the median,
while the area under the performance curves corresponds to the mean.
</p>
<p>
To demonstrate the superiority of our methods over the baselines and provide reliable evaluation results,
we follow the
evaluation protocols proposed in <a href="https://arxiv.org/abs/2108.13264" target="_blank">(Agarwal et
al., 2021)</a>.
Specifically, we adopt the
evaluation methods for all methods with N tasks × N seeds runs in total.
Evaluation results demonstrate that our method outperforms the baseline methods by asignificant margin
based
on all four reliable metrics.
</p>
</div>
</div>
</div>
<!--/ Concurrent Work. -->
<div class="container is-max-desktop">
<!-- Concurrent Work. -->
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Related Work</h2>
<div class="content has-text-justified">
<p>
This project would not be possible without the following wonderful prior work.
</p>
<p>
<a href="https://github.com/microsoft/oac-explore">Optimistic Actor Critic</a> gave inspiration to our
method,
<a href="https://github.com/Farama-Foundation/D4RL">D4RL</a>
provides the dataset and benchmark for evaluating the performance of our agent, and
<a href="https://github.com/rail-berkeley/rlkit/">RLkit</a> offered a strong RL framework
for building our code from.
</p>
</div>
</div>
</div>
<!--/ Concurrent Work. -->
</section>
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>@misc{li2022offline,
title={Offline Reinforcement Learning with Closed-Form Policy Improvement Operators},
author={Jiachen Li and Edwin Zhang and Ming Yin and Qinxun Bai and Yu-Xiang Wang and William Yang Wang},
journal={ICML},
year={2023},
}</code></pre>
</div>
</section>
<footer class="footer">
<div class="container">
<div class="content has-text-centered">
<a class="icon-link" href="https://proceedings.mlr.press/v202/li23av/li23av.pdf">
<i class="fas fa-file-pdf"></i>
</a>
<a class="icon-link" href="https://github.com/cfpo-icml23/cfpi" class="external-link" disabled>
<i class="fab fa-github"></i>
</a>
</div>
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p style="text-align: center
">
Website templated borrowed from <a href="https://github.com/nerfies/nerfies.github.io">nerfies.</a>
</p>
</div>
</div>
</div>
</div>
</footer>
</body>
</html>