-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathreservoirs3.html
331 lines (289 loc) · 9.79 KB
/
reservoirs3.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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta http-equiv="X-UA-Compatible" content="IE=edge">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Reservoir Computing</title>
<style>
body {
font-family: Arial, sans-serif;
background-color: #f0f0f0;
margin: 0;
padding: 0;
}
.header {
text-align: center;
padding: 20px;
}
.header img {
width: 200px;
height: 200px;
border-radius: 50%;
}
.header h1 {
margin-top: 10px;
margin-bottom: 0;
}
.header p {
margin-top: 0;
}
.nav {
text-align: center;
background-color: #333;
padding: 10px;
}
.nav a {
color: #fff;
text-decoration: none;
padding: 0 10px;
}
.nav a:hover {
text-decoration: underline;
}
.content {
padding: 20px;
}
.content {
padding: 20px;
background-color: #ffffff; /* this is the color white */
}
body {
font-family: Arial, sans-serif;
background-color: #f0f0f0;
margin: 0;
padding: 0;
}
.section {
display: flex;
align-items: center;
justify-content: space-around;
margin: 30px auto;
padding: 40px;
background-color: #fff;
border-radius: 5px;
box-shadow: 0 0 15px rgba(0,0,0,0.1);
width: 90%;
max-width: 1200px;
}
.section .img1 {
width: 100px;
height: 150px;
}
.section .img2 {
width: 450px;
height: 200px;
}
.section .img3 {
width: 500px;
height: 300px;
}
.section .img4 {
width: 100px;
height: 100px;
}
.section div {
margin-left: 40px;
width: 70%;
}
.section:nth-child(odd) div {
margin-right: 40px;
margin-left: 0;
}
.section:nth-child(odd) {
flex-direction: row-reverse;
}
.intro {
display: flex;
/*align-items: center;*/
/*justify-content: center;*/
margin: 30px auto;
padding: 30px;
background-color: #333;
color: #fff;
border-radius: 5px;
width: 90%;
max-width: 1200px;
/*text-align: center;*/
}
.intro a {
color: #FFD700; /* Change the color of links */
text-decoration: none; /* Remove the underline */
padding: 0px; /* Add some space around the links */
}
.intro a:hover {
color: #fff; /* Change the color when hovered */
text-decoration: underline; /* Underline when hovered */
}
.intro a:first-child {
padding-left: 0;
padding-right: 0;
}
/* CSS for Screen Width up to 600px (Mobile) */
@media only screen and (max-width: 600px) {
.header img {
width: 100px;
height: 100px;
}
.section {
flex-direction: column;
align-items: center;
padding: 20px;
width: 100%;
}
.section .img1, .section .img2, .section .img3, .section .img4 {
width: 100px;
height: auto;
}
.section div {
margin-left: 0;
margin-right: 0;
}
.intro {
flex-direction: column;
align-items: center;
padding: 20px;
width: 100%;
}
}
/* CSS for Screen Width up to 600px (Mobile) */
@media only screen and (max-width: 600px) {
.header img {
width: 100px;
height: 100px;
}
.section {
flex-direction: column;
align-items: center;
padding: 20px;
width: 100%;
}
.section .img1, .section .img2, .section .img3 {
width: 100px;
height: auto;
}
.section div {
margin-left: 0;
margin-right: 0;
}
.intro {
flex-direction: column;
align-items: center;
padding: 20px;
width: 100%;
}
}
.section .img1, .section .img2, .section .img3, .section .img4 {
width: 40%;
height: auto;
}
h1, h2, h4, h5, h6 {
text-align: center;
}
.image-background .esnimg {
background-color: #000;
display: flex;
justify-content: center;
align-items: center;
padding: 0px;
margin: auto;
}
.image-background .esnimg {
max-width: 50%;
height: auto;
}
/* Add styles for the .content paragraph */
.content p {
max-width: 800px;
margin: auto;
padding: 15px;
background-color: rgba(0,0,0,5%);
border-radius: 3px;
}
.content article {
max-width: 900px;
margin: auto;
}
/* Add this code */
.content {
max-width: 1000px;
margin: 0 auto;
}
.content p {
background-color: #f2f2f2; /* Pale gray */
padding: 15px;
border-radius: 5px;
}
.content .paragraph {
background-color: #f2f2f2; /* Pale gray */
padding: 15px;
border-radius: 5px;
margin-bottom: 15px; /* gives some space between paragraphs */
max-width: 800px;
margin-left: auto;
margin-right: auto;
}
.content .paragraph p {
margin: 0;
}
</style>
<link rel="stylesheet" href="style.css">
</head>
<body>
<div class="header">
<img src="portrait.jpg" alt="Your headshot">
<h1>Anthony Polloreno, Ph.D.</h1>
<h2>Research Engineer</h2>
</div>
<div class="nav">
<a href="index.html">Home</a> |
<a href="research.html">Research</a> |
<a href="engineering.html"> Engineering </a> |
<a href="https://twitter.com/ampolloreno">Twitter</a> |
<a href="https://www.linkedin.com/in/ampolloreno/">LinkedIn</a> |
<a href="https://github.com/ampolloreno">GitHub</a> |
<a href="papers.html">Papers</a> |
<a href="resume.pdf">Resume</a>|
<a href="reservoirs.html">Reservoir Computing I</a> |
<a href="reservoirs2.html">Reservoir Computing II</a>|
<a href="reservoirs3.html">Reservoir Computing III</a>
</div>
<div class="content" id="blogPost">
<article>
<h1>The Impact of Noise on Recurrent Neural Networks III</h1>
<div class="paragraph">
<p> We are finally set to analyze the impacts of noise on our particular model of recurrent computation - reservoir computing
with echo state networks. In the <a href="reservoirs2.html">previous post</a>, we implemented a GPU simulator together with
an extremely simple noise model by adding Gaussian noise to each element of the output signal from the reservoir.
In principle, the dynamics of a system will have more complicated noise based on the details of the computation being done,
but in our case using such a simple model will let us explore intuitively why we should expect a substantial degradation in
performance in the first place. Based on our discussion in the first post, we have additionally considered all products of the
output signals, and are assessing the performance of the reservoir by using it for the NARMA10 task.
</p>
<p>
In particular, we are going to analyze the impact of the Gaussian noise by considering a simple model of how the signals
become corrupted. We naively expect an exponential number of product signals because the number of possible product signals
from a set of signals is given by the cardinality of its power set - there are an exponential number of possible combinations
from any set. We will group these product signals by the number of terms in each product, and compute the probability
that the product signal is uncorrupted from the noise. Doing this, we will be able to reason about the expected number
of useful signals, and establish that with noise the expected number of signals should scale polynomially, rather than exponentially.
</p>
<p>
Once we have established that the expected number of signals in this simple model is approximately a monomial in the system size,
rather than exponential, we will plot how we expect the degree of the monomial to vary with noise. To think about
the limits of this plot - when the noise becomes very large, the echo state network will be unable to learn, and hence
their performance will be independent of the number of signals, giving a degree of zero. When the noise goes to zero,
then we can expect the reservoir will be able to make use of the exponentially-many signals, and the approximation
of a monomial dependence becomes invalid (the monomial degree becomes arbitrarily large). Let's give the code a look.
</p>
</div>
<iframe src="Webpost3.html" style="height:8000px; width:100%; border:none;" ></iframe>
<div class="paragraph">
<section style="text-align: center;">
<h3>Acknowledgements</h3>
<p>A special thanks to <a href="https://ohaithe.re/">Alex Meiburg</a>, <a href="https://andremelo.org/">André Melo</a> and <a href="https://chromotopy.org/">Eric Peterson</a> for feedback on this post!</p>
</section>
</article>
</div>
</body>
</html>