-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathlog.html
396 lines (396 loc) · 24.9 KB
/
log.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
<!-- <div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">May 5, 2020</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--bayesian-optimization/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2020/bayesian-optimization">
<div class="thumbnail"><img src="2020/bayesian-optimization/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Exploring Bayesian Optimization</h2>
<p class="authors">Apoorv Agnihotri and Nipun Batra</p>
<p class="abstract">How to tune hyperparameters for your machine learning model using Bayesian optimization.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">March 16, 2020</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--grand-tour/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2020/grand-tour">
<div class="thumbnail"><img src="2020/grand-tour/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Visualizing Neural Networks with the Grand Tour</h2>
<p class="authors">Mingwei Li, Zhenge Zhao, and Carlos Scheidegger</p>
<p class="abstract">By focusing on linear dimensionality reduction, we show how to visualize many dynamic phenomena in neural networks.</p>
</div>
</a>
</div>
<div class="post-preview thread">
<div class="metadata">
<div class="publishedDate">March 10, 2020</div>
<div class="tags">
<span class="tag thread">Thread</span>
</div>
</div>
<a href="2020/circuits">
<div class="thumbnail"><img src="2020/circuits/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Thread: Circuits</h2>
<p class="authors">Nick Cammarata, Shan Carter, Gabriel Goh, Chris Olah, Michael Petrov, and Ludwig Schubert</p>
<p class="abstract">What can we learn if we invest heavily in reverse engineering a single neural network?</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">Jan. 10, 2020</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--attribution-baselines/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2020/attribution-baselines">
<div class="thumbnail"><img src="2020/attribution-baselines/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Visualizing the Impact of Feature Attribution Baselines</h2>
<p class="authors">Pascal Sturmfels, Scott Lundberg, and Su-In Lee</p>
<p class="abstract">Exploring the baseline input hyperparameter, and how it impacts interpretations of neural network behavior.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">Nov. 4, 2019</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--receptive-field/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2019/computing-receptive-fields">
<div class="thumbnail"><img src="2019/computing-receptive-fields/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Computing Receptive Fields of Convolutional Neural Networks</h2>
<p class="authors">André Araujo, Wade Norris, and Jack Sim</p>
<p class="abstract">Detailed derivations and open-source code to analyze the receptive fields of convnets.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">Sept. 30, 2019</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--td-paths/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2019/paths-perspective-on-value-learning">
<div class="thumbnail"><img src="2019/paths-perspective-on-value-learning/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">The Paths Perspective on Value Learning</h2>
<p class="authors">Sam Greydanus and Chris Olah</p>
<p class="abstract">A closer look at how Temporal Difference Learning merges paths of experience for greater statistical efficiency</p>
</div>
</a>
</div>
<div class="post-preview commentary">
<div class="metadata">
<div class="publishedDate">Aug. 6, 2019</div>
<div class="tags">
<span class="tag commentary">Commentary</span>
</div>
</div>
<a href="2019/advex-bugs-discussion">
<div class="thumbnail"><img src="2019/advex-bugs-discussion/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">A Discussion of ‘Adversarial Examples Are Not Bugs, They Are Features’</h2>
<p class="authors">Logan Engstrom, Justin Gilmer, Gabriel Goh, Dan Hendrycks, Andrew Ilyas, Aleksander Madry, Reiichiro Nakano, Preetum Nakkiran, Shibani Santurkar, Brandon Tran, Dimitris Tsipras, and Eric Wallace</p>
<p class="abstract">Six comments from the community and responses from the original authors</p>
</div>
</a>
</div>
<div class="post-preview commentary">
<div class="metadata">
<div class="publishedDate">April 9, 2019</div>
<div class="tags">
<span class="tag commentary">Commentary</span>
</div>
</div>
<a href="2019/gan-open-problems">
<div class="thumbnail"><img src="2019/gan-open-problems/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Open Questions about Generative Adversarial Networks</h2>
<p class="authors">Augustus Odena</p>
<p class="abstract">What we’d like to find out about GANs that we don’t know yet.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">April 2, 2019</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--visual-exploration-gaussian-processes/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2019/visual-exploration-gaussian-processes">
<div class="thumbnail"><img src="2019/visual-exploration-gaussian-processes/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">A Visual Exploration of Gaussian Processes</h2>
<p class="authors">Jochen Görtler, Rebecca Kehlbeck, and Oliver Deussen</p>
<p class="abstract">How to turn a collection of small building blocks into a versatile tool for solving regression problems.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">March 25, 2019</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--memorization-in-rnns/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2019/memorization-in-rnns">
<div class="thumbnail"><img src="2019/memorization-in-rnns/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Visualizing memorization in RNNs</h2>
<p class="authors">Andreas Madsen</p>
<p class="abstract">Inspecting gradient magnitudes in context can be a powerful tool to see when recurrent units use short-term or long-term contextual understanding.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">March 6, 2019</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--activation-atlas/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2019/activation-atlas">
<div class="thumbnail"><img src="2019/activation-atlas/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Activation Atlas</h2>
<p class="authors">Shan Carter, Zan Armstrong, Ludwig Schubert, Ian Johnson, and Chris Olah</p>
<p class="abstract">By using feature inversion to visualize millions of activations from an image classification network, we create an explorable activation atlas of features the network has learned and what concepts it typically represents.</p>
</div>
</a>
</div>
<div class="post-preview commentary">
<div class="metadata">
<div class="publishedDate">Feb. 19, 2019</div>
<div class="tags">
<span class="tag commentary">Commentary</span>
</div>
</div>
<a href="2019/safety-needs-social-scientists">
<div class="thumbnail"><img src="2019/safety-needs-social-scientists/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">AI Safety Needs Social Scientists</h2>
<p class="authors">Geoffrey Irving and Amanda Askell</p>
<p class="abstract">If we want to train AI to do what humans want, we need to study humans.</p>
</div>
</a>
</div>
<div class="post-preview editorial">
<div class="metadata">
<div class="publishedDate">Aug. 14, 2018</div>
<div class="tags">
<span class="tag editorial">Editorial</span>
</div>
</div>
<a href="2018/editorial-update">
<div class="thumbnail"><img src="2018/editorial-update/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Distill Update 2018</h2>
<p class="authors">Distill Editors</p>
<p class="abstract">An Update from the Editorial Team</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">July 25, 2018</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--differentiable-parameterizations/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2018/differentiable-parameterizations">
<div class="thumbnail"><img src="2018/differentiable-parameterizations/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Differentiable Image Parameterizations</h2>
<p class="authors">Alexander Mordvintsev, Nicola Pezzotti, Ludwig Schubert, and Chris Olah</p>
<p class="abstract">A powerful, under-explored tool for neural network visualizations and art.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">July 9, 2018</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--feature-wise-transformations/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2018/feature-wise-transformations">
<div class="thumbnail"><img src="2018/feature-wise-transformations/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Feature-wise transformations</h2>
<p class="authors">Vincent Dumoulin, Ethan Perez, Nathan Schucher, Florian Strub, Harm de Vries, Aaron Courville, and Yoshua Bengio</p>
<p class="abstract">A simple and surprisingly effective family of conditioning mechanisms.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">March 6, 2018</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--building-blocks/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2018/building-blocks">
<div class="thumbnail"><img src="2018/building-blocks/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">The Building Blocks of Interpretability</h2>
<p class="authors">Chris Olah, Arvind Satyanarayan, Ian Johnson, Shan Carter, Ludwig Schubert, Katherine Ye, and Alexander Mordvintsev</p>
<p class="abstract">Interpretability techniques are normally studied in isolation. We explore the powerful interfaces that arise when you combine them — and the rich structure of this combinatorial space.</p>
</div>
</a>
</div>
<div class="post-preview commentary">
<div class="metadata">
<div class="publishedDate">Dec. 4, 2017</div>
<div class="tags">
<span class="tag commentary">Commentary</span>
</div>
</div>
<a href="2017/aia">
<div class="thumbnail"><img src="2017/aia/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Using Artificial Intelligence to Augment Human Intelligence</h2>
<p class="authors">Shan Carter and Michael Nielsen</p>
<p class="abstract">By creating user interfaces which let us work with the representations inside machine learning models, we can give people new tools for reasoning.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">Nov. 27, 2017</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--ctc/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2017/ctc">
<div class="thumbnail"><img src="2017/ctc/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Sequence Modeling with CTC</h2>
<p class="authors">Awni Hannun</p>
<p class="abstract">A visual guide to Connectionist Temporal Classification, an algorithm used to train deep neural networks in speech recognition, handwriting recognition and other sequence problems.</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">Nov. 7, 2017</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--feature-visualization/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2017/feature-visualization">
<div class="thumbnail"><img src="2017/feature-visualization/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Feature Visualization</h2>
<p class="authors">Chris Olah, Alexander Mordvintsev, and Ludwig Schubert</p>
<p class="abstract">How neural networks build up their understanding of images</p>
</div>
</a>
</div>
<div class="post-preview peer-reviewed">
<div class="metadata">
<div class="publishedDate">April 4, 2017</div>
<div class="tags">
<a class="tag peer-reviewed" title="View this article's reviews as Github issues." href="https://github.com/distillpub/post--momentum/issues?q=is%3Aissue+label%3Apeer-review"><span>Peer-reviewed</span></a>
</div>
</div>
<a href="2017/momentum">
<div class="thumbnail"><img src="2017/momentum/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Why Momentum Really Works</h2>
<p class="authors">Gabriel Goh</p>
<p class="abstract">We often think of optimization with momentum as a ball rolling down a hill. This isn’t wrong, but there is much more to the story.</p>
</div>
</a>
</div>
<div class="post-preview commentary">
<div class="metadata">
<div class="publishedDate">March 22, 2017</div>
<div class="tags">
<span class="tag commentary">Commentary</span>
</div>
</div>
<a href="2017/research-debt">
<div class="thumbnail"><img src="2017/research-debt/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Research Debt</h2>
<p class="authors">Chris Olah and Shan Carter</p>
<p class="abstract">Science is a human activity. When we fail to distill and explain research, we accumulate a kind of debt...</p>
</div>
</a>
</div>
<div class="post-preview ">
<div class="metadata">
<div class="publishedDate">Dec 6, 2016</div>
<div class="tags">
</div>
</div>
<a href="2016/handwriting">
<div class="thumbnail"><img src="2016/handwriting/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Experiments in Handwriting with a Neural Network</h2>
<p class="authors">Shan Carter, David Ha, Ian Johnson, and Chris Olah</p>
<p class="abstract">Several interactive visualizations of a generative model of handwriting. Some are fun, some are serious.</p>
</div>
</a>
</div>
<div class="post-preview ">
<div class="metadata">
<div class="publishedDate">Oct 17, 2016</div>
<div class="tags">
</div>
</div>
<a href="2016/deconv-checkerboard">
<div class="thumbnail"><img src="2016/deconv-checkerboard/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Deconvolution and Checkerboard Artifacts</h2>
<p class="authors">Augustus Odena, Vincent Dumoulin, and Chris Olah</p>
<p class="abstract">When we look very closely at images generated by neural networks, we often see a strange checkerboard pattern of artifacts.</p>
</div>
</a>
</div>
<div class="post-preview ">
<div class="metadata">
<div class="publishedDate">Oct 13, 2016</div>
<div class="tags">
</div>
</div>
<a href="2016/misread-tsne">
<div class="thumbnail"><img src="2016/misread-tsne/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">How to Use t-SNE Effectively</h2>
<p class="authors">Martin Wattenberg, Fernanda Viégas, and Ian Johnson</p>
<p class="abstract">Although extremely useful for visualizing high-dimensional data, t-SNE plots can sometimes be mysterious or misleading.</p>
</div>
</a>
</div>
<div class="post-preview ">
<div class="metadata">
<div class="publishedDate">Sept 8, 2016</div>
<div class="tags">
</div>
</div>
<a href="2016/augmented-rnns">
<div class="thumbnail"><img src="2016/augmented-rnns/thumbnail.jpg"></div>
<div class="description">
<h2 class="title">Attention and Augmented Recurrent Neural Networks</h2>
<p class="authors">Chris Olah and Shan Carter</p>
<p class="abstract">A visual overview of neural attention, and the powerful extensions of neural networks being built on top of it.</p>
</div>
</a>
</div> -->