-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
188 lines (165 loc) · 8.89 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
<!DOCTYPE HTML>
<html>
<head>
<style>
.video-container {
width: 240px;
height: 240px;
overflow: hidden;
position: relative;
}
.video-container video {
position: absolute;
top: 0;
left: 0;
width: 100%;
height: 100%;
object-fit: cover;
}
body {
font-family: Arial, sans-serif;
background-color: #f8f8f8;
color: #404040;
/* padding: 0 25%; Adjust as needed */
}
.content-section {
padding: 0 25%; /* Add padding to content sections only */
}
h1 {
color: #800000;
}
h2 {
color: #404040;
}
table {
width: 100%;
}
td, th {
padding: 15px;
text-align: left;
}
/* Modify only the first column of the second table */
#secondTable td:first-child, #secondTable th:first-child {
width: 5%; /* Adjust as needed */
}
#secondTable {
padding: 0 15%; /* Adjust as needed */
}
#secondTable td, #secondTable th {
border: 1px solid black; /* This will give a border to your table cells */
padding: 15px;
text-align: left;
}
hr {
border-top: 2px solid #4a3737;
}
video {
width: 100%;
height: auto;
}
</style>
</head>
<body onload="init()">
<div id="main">
<div class="content-section">
<h1>Effective and Efficient Reinforcement Learning Based on Structural Information Principles (SIDM)</h1>
<div id="authors">
<table>
<tr>
<td><a href="https://scholar.google.com/citations?hl=en&user=PZ3jhoMAAAAJ">Xianghua Zeng</a></td>
<td><a href="https://scholar.google.com/citations?user=R25rbyQAAAAJ&hl=en&oi=ao">Hao Peng</a></td>
<td>Dingli Su</td>
<td><a href="https://dblp.org/pid/66/4917.html">Angsheng Li</a></td>
</tr>
<tr>
<td class="affiliation">School of Computer Science and Engineering, Beihang University</td>
<td class="affiliation">School of Cyber Science and Technology, Beihang University</td>
<td class="affiliation">School of Computer Science and Engineering, Beihang University</td>
<td class="affiliation">School of Computer Science and Engineering, Beihang University</td>
</tr>
</table>
</div>
<hr>
<p>
<b>TL;DR</b> We propose a new, unsupervised, and adaptive Decision-Making framework called SIDM for Reinforcement Learning. This approach handles high complexity environments without manual intervention, and increases sample efficiency and policy effectiveness.
</p>
<p>
<a href="link_to_your_paper">Paper</a> |
<a href="https://github.com/RingBDStack/SIDM">Code</a>
</p>
<hr>
<div id="abstract">
<p>
<b>Abstract</b> Although Reinforcement Learning (RL) algorithms acquire sequential behavioral patterns through interactions with the environment, their effectiveness in noisy and high-dimensional scenarios typically relies on specific structural priors. This paper proposes an unsupervised and adaptive Decision-Making framework called SIDM for RL, which uses action and state abstractions to address this issue. SIDM improves policy quality, stability, and sample efficiency by up to 32.70%, 88.26%, and 64.86%, respectively.
</p>
</div>
<hr>
<div id="approach">
<img src="./docs/framework.png" alt="Framework" style="display: block; max-width: 100%; margin: auto;">
<p>
<b>Approach</b> The SIDM processes environmental observations and rewards, retains historical trajectories, and outputs state, abstract action, role set, and skill set to specific downstream algorithms. Initially, we employ encoder-decoder architectures to embed environmental observations and actions, measure feature similarities and eliminate trivial edges to construct state and action graphs. Subsequently, we initialize an encoding tree for each graph, minimize their structural entropy to obtain community partitioning for states and actions, and design an aggregation using assigned entropy as weights to achieve hierarchical abstractions. Finally, we extract abstract elements to construct a transition graph, define and optimize the structural entropy for this directed graph, calculate common path entropy to quantify each transition's sample probability, and introduce an adaptive skill-based learning mechanism.
</p>
</div>
<hr>
<p>
<a name="videos"/>
<b>Videos</b>
</div>
</p>
<table id="secondTable">
<tr>
<th>Task</th>
<th>SIDM</th>
<th>SAC</th>
<th>SE</th>
<th>HIRO</th>
<th>HSD</th>
</tr>
<tr>
<td>Hurdles</td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SISL-hurdles-49.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SAC-hurdles-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SE-hurdles-80.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HIRO-hurdles-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HSD-hurdles-9.mp4" type="video/mp4"></video></td>
</tr>
<tr>
<td>Limbo</td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SISL-limbo-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SAC-limbo-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SE-limbo-9.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HIRO-limbo-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HSD-limbo-0.mp4" type="video/mp4"></video></td>
</tr>
<tr>
<td>Pole
Balance</td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SISL-pbalance-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SAC-pbalance-40.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SE-pbalance-77.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HIRO-pbalance-123.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HSD-pbalance-8.mp4" type="video/mp4"></video></td>
</tr>
<tr>
<td>Hurdles
Limbo</td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SISL-hurdleslimbo-52.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SAC-hurdleslimbo-111.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SE-hurdleslimbo-60.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HIRO-hurdleslimbo-100.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HSD-hurdleslimbo.mp4" type="video/mp4"></video></td>
</tr>
<tr>
<td>Stairs</td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SISL-stairs-53.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SAC-stairs-160.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/SE-stairs-82.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HIRO-stairs-188.mp4" type="video/mp4"></video></td>
<td class="video-container"><video width="320" height="240" controls autoplay muted><source src="./video/HSD-stairs-34.mp4" type="video/mp4"></video></td>
</tr>
<!-- Additional rows as needed -->
</table>
<hr>
</div>
</body>
</html>