-
Notifications
You must be signed in to change notification settings - Fork 1
/
2018.html
470 lines (407 loc) · 19.9 KB
/
2018.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
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>1st VISxAI Workshop at IEEE VIS 2018</title>
<link rel="stylesheet"
href="node_modules/bootswatch/dist/sandstone/bootstrap.css">
<link rel="stylesheet"
href="styles.css">
<link href="https://fonts.googleapis.com/css?family=IBM+Plex+Mono|IBM+Plex+Sans"
rel="stylesheet">
<link rel="icon" type="image/png" sizes="32x32" href="favicon-32x32.png">
<link rel="icon" type="image/png" sizes="96x96" href="favicon-96x96.png">
<link rel="icon" type="image/png" sizes="16x16" href="favicon-16x16.png">
<script
src="https://code.jquery.com/jquery-3.3.1.slim.min.js"
integrity="sha256-3edrmyuQ0w65f8gfBsqowzjJe2iM6n0nKciPUp8y+7E="
crossorigin="anonymous"></script>
<script src="node_modules/bootstrap/js/dist/util.js"></script>
<script src="node_modules/bootstrap/js/dist/collapse.js"></script>
<!-- Share card -->
<meta name="twitter:card" content="summary" />
<meta name="twitter:site" content="@visxai" />
<meta name="twitter:creator" content="@visxai" />
<meta property="og:url" content="http://visxai.io" />
<meta property="og:title" content="Workshop on Visualization for AI Explainability" />
<meta property="og:description"
content="The role of visualization in artificial intelligence (AI) gained significant attention in recent years. With the growing complexity of AI models, the critical need for understanding their inner-workings has increased. Visualization is potentially a powerful technique to fill such a critical need. The goal of this workshop is to initiate a call for 'explainables' / 'explorables' that explain how AI techniques work using visualization. We believe the VIS community can leverage their expertise in creating visual narratives to bring new insight into the often obfuscated complexity of AI systems." />
<meta property="og:image" content="http://visxai.github.io/img/share.png" />
</head>
<body>
<div id="banner">
VISxAI is back! Join us at <a href="http://visxai.io">VISxAI 2019 at IEEE VIS</a> in Vancouver, Canada!
</div>
<nav class="navbar navbar-expand-lg navbar-light bg-light">
<a class="navbar-brand" href="/"><span class="vxa">VISxAI</span></a>
<button class="navbar-toggler" type="button" data-toggle="collapse"
data-target="#navbarColor03" aria-controls="navbarColor03"
aria-expanded="false" aria-label="Toggle navigation">
<span class="navbar-toggler-icon"></span>
</button>
<div class="collapse navbar-collapse" id="navbarColor03">
<ul class="navbar-nav mr-auto">
<li class="nav-item">
<a class="nav-link" href="submit.html">SUBMIT</a>
</li>
<li class="nav-item">
<a class="nav-link" href="#program">Program</a>
</li>
<li class="nav-item">
<a class="nav-link" href="#dates">Dates</a>
</li>
<li class="nav-item">
<a class="nav-link" href="#call">CFP</a>
</li>
<li class="nav-item">
<a class="nav-link" href="#orga">Organizers</a>
</li>
<li class="nav-item">
<a class="nav-link" href="#pc">PC</a>
</li>
</ul>
</div>
</nav>
<div class="container" style="margin-top: 12pt;">
<div class="float-right">
<!--<img src="img/logo.png" height="70"/>-->
<img src="img/logo_v2.png" height="70"/>
</div>
<!--<div style="position: absolute; right:5px;">-->
<!--</div>-->
<h2>1<sup>st</sup> Workshop on <br> <b>Visualization for AI Explainability</b></h2>
<p>October 22, 2018 at IEEE VIS in Berlin, Germany</p>
<p>
The role of visualization in artificial intelligence (AI) gained
significant attention in recent years. With the growing complexity of AI
models, the critical need for understanding their inner-workings has
increased. Visualization is potentially a powerful technique to fill
such a critical need.
</p>
<p>
The goal of this workshop is to initiate a call for “explainables” that
explain how AI techniques work using visualizations. We believe the VIS
community can leverage their expertise in creating visual narratives to
bring new insight into the often obfuscated complexity of AI systems.
</p>
<p class="text-center">
<img class="img-fluid" src="img/examples-2018.png">
<div class="figure-caption"> Examples of blog posts using interactive
visualizations to (1) explain general concepts of AI models and (2)
communicate experimental insights when playing with AImodels.
(a) <a href="https://distill.pub/2017/momentum/" target="_blank">Why
Momentum Really Works</a> by Gabriel Goh;
(b) <a href="http://playground.tensorflow.org/" target="_blank">Tensorflow
Playground</a> by Smilkov, Carter, et.al.;
(c) <a href="http://formafluens.io/client/mix10.html">FormaFluens Data
Experiment</a> by Strobelt, Phibbs, Martino.
</div>
</p>
<h2 id="program">Program Overview</h2>
<table style="padding: 5pt;">
<tr>
<td class="schedule">2:20 -- 2:25</td>
<td><b>Welcome from the Organizers</b></td>
</tr>
<tr>
<td class="schedule">2:25 -- 3:10</td>
<td><b>Keynote: Been Kim (Google Brain)</b>
<br><b>Towards Interpretability for Everyone: Testing with
Concept Activation Vectors (TCAV)</b>
<br>
The ultimate goal of interpretability is to help users gain
insights into the model for more responsible use of ML. Unlike
the majority of subfields in ML, interpretable ML requires
studying how humans parse complex information and exploring
effective ways to communicate such information. This human
aspect becomes even more critical when developing
interpretability methods for non-ML experts/layer users --- my
core research agenda. I will share some interpretability methods
that are designed with or without considering human aspect, and
where they succeeded or fall short. I will take a deeper dive in
one of my recent work - testing with concept activation vectors
(TCAV) - a post-training interpretability method for complex
models, such as neural network. This method provides an
interpretation of a neural net's internal state in terms of
human-friendly, high-level concepts instead of low-level input
features. Most importantly, I will share some open questions in
interpretability methods that are calling for visualization
community's expertise.
<!--"I am interested in designing high-performance machine learning methods that make sense to humans.-->
<!--My focus is building interpretability method for already-trained models or building inherently interpretable models.-->
<!--In particular, I believe the language of explanations should include higher-level, human-friendly concepts-->
<!--so that it can make sense to everyone."-->
</td>
</tr>
<tr>
<td class="schedule">3:10 -- 3:35</td>
<td><b>Session I: Neural Networks and Deep Learning </b><br>
<a href="http://eprints.uwe.ac.uk/37807/1/visualising_networks.html">Visualising
State Space Representations of Long Short-Term Memory
Networks</a> -- Emmanuel M. Smith, Jim Smith, Phil Legg and
Simon Francis<br>
<a href="http://hdc.cs.arizona.edu/~mwli/grandtour/index.html">Visualizing
neuron activations of neural networks with the grand
tour</a> -- Mingwei Li, Zhenge Zhao and Carlos
Scheidegger<br>
<a href="https://github.com/philippschmitt/embodied-ml">
Embodied Machine Learning: An educational, human MNIST
classifier</a> -- Philipp Schmitt<br>
</td>
</tr>
<tr>
<td class="schedule">3:35 -- 4:00</td>
<td><b>Session II: Projections and Dimensionality Reduction </b>
<br>
<a href="https://roadsfromabove.netlify.com/"> Roads from
Above</a> -- Greg More, Slaven Marusic and Caihao Cui <br>
<a href="https://idyll.pub/post/visxai-dimensionality-reduction-1dbad0a67a092b007c526a45/">
The Beginner's Guide to Dimensionality Reduction</a> --
Matthew Conlen and Fred Hohman<br>
<a href="https://visxprojections.dbvis.de/"> Dimension,
Distances, or Neighborhoods? Projection Literacy for the
Analysis of Multivariate Data </a> -- Dirk Streeb, Rebecca
Kehlbeck, Dominik Jäckle and Mennatallah El-Assady
</td>
</tr>
<tr>
<td class="schedule">4:00 -- 4:20</td>
<td><b>Coffee Break with Poster Session</b></td>
</tr>
<tr>
<td class="schedule">4:20 -- 4:45</td>
<td><b>Session III: Data Distribution and Bias </b><br>
<a href="https://www.jgoertler.com/visual-exploration-gaussian-processes/">
A Visual Exploration of Gaussian Processes</a> -- Jochen
Görtler, Rebecca Kehlbeck and Oliver Deussen <br>
<a href="https://spinthil.github.io/towards-an-interpretable-latent-space/">
Towards an Interpretable Latent Space </a> -- Thilo Spinner,
Jonas Körner, Jochen Görtler and Oliver Deussen <br>
<a href="https://mybinder.org/v2/gh/Jindong-Explainable-AI/Bias_in_Machine_Learning/master?filepath=ML_Bias.ipynb
"> Understanding Bias in Machine Learning </a> -- Jindong Gu and Daniela Oelke
</td>
</tr>
<tr>
<td class="schedule">4:45 -- 5:10</td>
<td><b>Session IV: Machine Learning Processes and Explanation
Strategies </b><br>
<a href="https://bib.dbvis.de/publications/details/787">Minions,
Sheep, and Fruits: Metaphorical Narratives to Explain
Artificial Intelligence and Build Trust</a> -- Wolfgang
Jentner, Rita Sevastjanova, Florian Stoffel, Daniel Keim, Jurgen
Bernard and Mennatallah El-Assady <br>
<a href="http://aimacode.github.io/aima-javascript/5-Adversarial-Search/">
Aimacode Javascript - Minimax</a> -- Michael Kawano<br>
<a href="http://verbalization.lingvis.io/"> Going beyond
Visualization: Verbalization as Complementary Medium to
Explain Machine Learning Models</a> -- Rita Sevastjanova,
Fabian Beck, Basil Ell, Cagatay Turkay, Rafael Henkin, Miriam
Butt, Daniel Keim and Mennatallah El-Assady
</td>
</tr>
<tr>
<td class="schedule">5:10 -- 5:55</td>
<td><b>Moderated Panel Discussion </b></td>
</tr>
<tr>
<td class="schedule">5:55 -- 6:00</td>
<td><b>Best submission ceremony and "Auf Wiedersehen" :)</b></td>
</tr>
<tr>
<td class="schedule">8:00 -- ...</td>
<td><b><a href="http://eastcoastparty.rocks">VISxAI Eastcoast
party </a></b></td>
</tr>
</table>
<br>
<h2>Posters</h2>
<a href="http://www.cs.williams.edu/~iris/res/bkt/">What is Bayesian
Knowledge Tracing?</a> -- Young Cho, Grace Mazzarella, Kelvin Tejeda,
Tongyu Zhou and Iris Howley <br>
<a href="https://www.bbvadata.com/recsys/">Recsys: what is a recommendation
in the Age of Machine Learning</a> -- Iskra Velitchkova, Juan Arévalo
and Marco Creatura<br>
<a href="https://sauln.github.io/blog/tda_explanations.html ">Understanding
ML through Topological Data Analysis </a> -- Nathaniel Saul and Dustin L
Arendt<br>
<a href="http://stelling.cc/machineplay/">Explaining neural network concepts
through an interactive visualization</a> -- Roberto Stelling and Adriana
S Vivacqua<br>
<a href="http://vis-server.win.tue.nl/visxai/">Plainability: Explainability
for 1-Dimensional Temporal Inputs</a> -- Humberto Simon Garcia
Caballero, Michel Westenberg and Binyam Gebre
<br>
<br>
<h2 id="awards">Awards</h2>
<strong>Best Paper</strong>
<br>
<a href="https://www.jgoertler.com/visual-exploration-gaussian-processes/">A Visual Exploration of Gaussian Processes</a> -- Jochen Görtler, Rebecca Kehlbeck and Oliver Deussen
<br>
<br>
<strong>Best Paper, Honorable Mention</strong>
<br>
<a href="https://idyll.pub/post/visxai-dimensionality-reduction-1dbad0a67a092b007c526a45/">The Beginner's Guide to Dimensionality Reduction</a> -- Matthew Conlen and Fred Hohman
<br>
<a href="https://roadsfromabove.netlify.com/">Roads from Above</a> -- Greg More, Slaven Marusic and Caihao Cui
<br>
<br>
<h2 id="dates">Important Dates</h2>
<pre>
July 12, 2018, 5:00pm PDT: Blog/Notebooks + Position Paper Submission
August 2, 2018: Author Notification
September 3, 2018: Camera-ready Copy for Accepted Submissions
September 7, 2018: VIS Early Bird Registration Ends
October 22 -- Workshop in Berlin at IEEE VIS 2018
</pre>
<h2 id="call">Call for Participation</h2>
<p> <strong>SUBMISSION CLOSED</strong></p>
<p>
To make our work more accessible to the general audience, we are
soliciting
submissions in a novel format: blog-style posts and jupyter-like
notebooks. In addition we also accept position papers in a more
traditional
form.
Please contact us, if you want to submit a original work in another
format. Email: <a href="mailto:[email protected]">orga.visxai at
gmail.com</a>
</p>
<h4>Explainables (Blogs, Markup, and Notebooks) </h4>
<p>Explainable submissions are the core element of the workshop, as this
workshop aims to be a platform for explanatory visualizations focusing
on AI
techniques.</p>
<p>Authors have the freedom to use whatever templates and formats they
like.
However, the narrative should be visual and interactive, and walk
readers
through a keen understanding on the ML technique or application. Authors
may
wish to write a <a href="https://distill.pub">Distill-style</a> blog
post (format),
interactive <a href="https://idyll-lang.org/">Idyll</a> markup,
or a <a href="http://jupyter.org">Jupyter</a> or <a
href="https://beta.observablehq.com/">Observable</a> notebook
that
integrates codes, visualizations to
tell the story.
</p>
<p>
Here are a few examples of visual explanations of AI methods in these types of formats:
<ul>
<li>[blog-style]
<a href="http://www.r2d3.us/visual-intro-to-machine-learning-part-1/"
target="_blank">http://www.r2d3.us/visual-intro-to-machine-learning-part-1/</a>
</li>
<li>[blog-style]
<a href="http://formafluens.io/client/mix10.html" target="_blank">http://formafluens.io/client/mix10.html</a>
</li>
<li>[markup]
<a href="https://idyll-lang.org/gallery/the-barnes-hut-approximation"
target="_blank">https://idyll-lang.org/gallery/the-barnes-hut-approximation</a>
</li>
<li>[notebook]
<a href="https://beta.observablehq.com/@nstrayer/t-sne-explained-in-plain-javascript"
target="_blank">https://beta.observablehq.com/@nstrayer/t-sne-explained-in-plain-javascript</a>
</li>
<li>[markup]
<a href="http://nbviewer.jupyter.org/github/agconti/kaggle-titanic/blob/master/Titanic.ipynb"
target="_blank">http://nbviewer.jupyter.org/github/agconti/kaggle-titanic/blob/master/Titanic.ipynb</a>
</li>
<li>[blog-style]
<a href="https://distill.pub/2017/momentum/" target="_blank">https://distill.pub/2017/momentum/</a>
</li>
</ul>
</p>
<p>While these examples are informative and excellent, we hope the
visualization community will think about ways to creatively expand on
such foundational work to explain AI methods using novel interactions
and visualizations often present at IEEE VIS.
Please contact us, if you want to submit a original work in another
format. Email: <a href="mailto:[email protected]">orga.visxai at
gmail.com</a>
</p>
<p>
The best works will be invited to submit their extended work to the
online publishing platform distill.pub to generate a cite-able
publication for authors.
</p>
<h4>Position Papers</h4>
<p>We will also accept position papers about impact and role of explainables
for VIS in AI. Submissions should be no more than 6 pages long and
formatted
according to the <a
href="http://junctionpublishing.org/vgtc/Tasks/camera.html">VGTC
formatting guidelines</a>.
Some good example of a position papers (not all for AI) can be found
here:
<ul>
<li><a href="https://graphics.cs.wisc.edu/Papers/2018/Gle18/viscomp.pdf"
target="_blank">Considerations for Visualizing Comparison</a>
</li>
<li>
<a href="https://bib.dbvis.de/uploadedFiles/RisktheDriftStretchingDisciplinaryBoundariesthroughCriticalCollaborationsbetweentheHumanitiesandVisualization.pdf"
target="_blank">Risk the Drift! Stretching Disciplinary
Boundaries through Critical Collaborations between the
Humanities and Visualization</a></li>
<li>
<a href="https://www.sciencedirect.com/science/article/pii/S2468502X17300086"
target="_blank">Towards better analysis of machine learning
models: A visual analytics perspective</a></li>
<li>
<a href="http://www.sci.utah.edu/publications/Che2017a/Theoretical-Advances-Visualization-2017.pdf"
target="_blank">Pathways for Theoretical Advances in
Visualization</a></li>
</ul>
</p>
<h2 id="orga">Organizers <span style="font-size: small">(alphabetic)</span>
</h2>
<p>
Mennatallah El-Assady - University of Konstanz<br/>
Duen Horng (Polo) Chau - Georgia Tech<br/>
Adam Perer - Carnegie Mellon University<br/>
Hendrik Strobelt - IBM Research, MIT-IBM Watson AI Lab<br/>
Fernanda Viégas - Google Brain
</p>
<h2 id="pc">Program Committee</h2>
Adam Perer<br>
Alexander Rush<br>
Arvind Satyanarayan<br>
Brady Redfearn<br>
Carlos Scheidegger<br>
Jaegul Choo<br>
Christian Bors<br>
Christopher Collins<br>
David Bau<br>
Duen Horng (Polo) Chau<br>
Dustin Arendt<br>
Dylan Cashman<br>
Lana El Sanyoura<br>
Fernanda Viégas<br>
Fred Hohman<br>
Hendrik Strobelt<br>
Iris Howley<br>
Juergen Bernard<br>
Kanit Wongsuphasawat<br>
Martin Wattenberg<br>
Matthew Conlen<br>
Mennatallah El-Assady<br>
Minsuk Kahng<br>
Rita Borgo<br>
Sebastian Gehrmann<br>
Tommy Dang<br>
Yamini Bansal<br>
Yang Wang<br>
<p> </p>
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async
src="https://www.googletagmanager.com/gtag/js?id=UA-119596896-1"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() {dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-119596896-1');
</script>
</div>
</body>
</html>