Skip to content

Commit 7301f20

Browse files
dontcallmedomdontcallmedom
committed
Publish minutes from last session
Also update previous minutes to incorporate related slides in the record Add horizontal navigation across minutes
1 parent e30e74c commit 7301f20

10 files changed

+545
-71
lines changed

‎_includes/horizontal-nav.html

+6-6
Original file line numberDiff line numberDiff line change
@@ -1,13 +1,13 @@
11
<p class=buttons>
22
{% if page.prevlink %}
3-
<button type=submit form=form formaction="{{page.prevlink}}"
4-
class="nav picto im-arrow-left">Previous: {{page.prevtitle}}</button>
3+
<a href="{{page.prevlink}}"
4+
class="nav picto im-arrow-left">Previous: {{page.prevtitle}}</a>
55
{% endif %}
6-
<a href="../presentations.html" class="picto im-data nav"
7-
>All talks</a>
6+
<a href="{% if page.uplink %}{{page.uplink}}{% else %}../presentations.html{% endif %}" class="picto im-data nav"
7+
>{% if page.uplink %}{{page.uptitle}}{% else %}All talks{% endif %}</a>
88
{% if page.nextlink %}
9-
<button class=nav type=submit form=form formaction="{{page.nextlink}}">Next:
9+
<a class=nav href="{{page.nextlink}}">Next:
1010
{{page.nexttitle}}<span
11-
class="picto im-arrow-right"></span></button>
11+
class="picto im-arrow-right"></span></a>
1212
{% endif %}
1313
</p>

‎_layouts/subpage.html

+11
Original file line numberDiff line numberDiff line change
@@ -15,6 +15,9 @@
1515
"Bringing together experts to enrich the Open Web Platform with better foundations for machine learning.">
1616
<meta property="og:image" content=
1717
"https://www.w3.org/2020/06/machine-learning-workshop/images/banner.png">
18+
{% if page.hasPDF %}
19+
<link rel="stylesheet" href="https://unpkg.com/[email protected]/web/pdf_viewer.css">
20+
{% endif %}
1821
</head>
1922
<body>
2023
<header class="header">
@@ -50,6 +53,9 @@ <h2>
5053
</ul>
5154
</nav>
5255
<h2>{{title}}</h2>
56+
{% if page.uplink %}
57+
{% include horizontal-nav.html %}
58+
{% endif %}
5359
</header>
5460
{% include sponsors.html %}
5561
{{ content }}
@@ -72,6 +78,11 @@ <h2>{{title}}</h2>
7278
"mailto:[email protected]">[email protected]</a>&gt;.
7379
</p>
7480
</footer>
81+
{% if page.hasPDF %}
82+
<script src="https://unpkg.com/[email protected]/build/pdf.js"></script>
83+
<script src="https://unpkg.com/[email protected]/web/pdf_viewer.js"></script>
84+
<script src="{{page.base}}pdf-loader.js"></script>
85+
{% endif %}
7586
<script src="{{page.base }}script.js">
7687
</script>
7788
<script src="{{page.base }}submit.js">

‎minutes/20200916.html

+27-7
Original file line numberDiff line numberDiff line change
@@ -3,9 +3,15 @@
33
layout: subpage
44
base: ../
55
parent: proceeding.html
6+
hasPDF: true
7+
nextlink: 20200922.html
8+
nexttitle: "Minutes of 2nd Live Session (September 22, 2020)"
9+
uplink: ../proceeding.html
10+
uptitle: List of minutes
611
---
712
<h2>Minutes of 1st Live Session (September 16 2020)</h2>
813
<address>Scribe: François Daoust</address>
14+
<p><strong><a href="https://github.com/w3c/machine-learning-workshop/tree/master/agenda">Agenda</a></strong> - <strong><a href="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf#page=1">Slides</a></strong><p>
915
<p><cite>Present:</cite> Dominique Hazael-Massieux, Anssi
1016
Kostiainen, Bernard Aboba, Deng Yuguang, Gabriel Vigliensoni, Jason
1117
Mayes, Jeff Hammond, Jun Weifu, Qiang Chen, Sangwhan Moon, Wendy
@@ -26,25 +32,30 @@ <h2>Minutes of 1st Live Session (September 16 2020)</h2>
2632
Rachel Yager, Mark Crawford, Piotr Migdal, Chenzelun, Marie-Claire
2733
Forgue, John Rochford, Virginie Galindo,</p>
2834
<h2>Introduction to the workshop</h2>
35+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-2"></div>
2936
<p><cite>Anssi:</cite> Welcome. I work with Intel, chair of the
3037
workshop. Background in Web standards and Web technologies. The
3138
purpose of this workshop, in a nutshell, is to bring together
3239
machine learning and web people to make sure that the Web platform
3340
has better foundations for machine learning. We want to understand
3441
how in-browser machine learning fits into the machine learning
3542
ecosystem, and explore things both ways. Many pieces of the
36-
platform could be improved.</p>
43+
platform could be improved.</p>
44+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-3"></div>
3745
<p><cite>Anssi:</cite> Why are we here? The reality is that people
3846
from diverse backgrounds need to gather together to complement each
39-
other, and better evolve the Web platform.</p>
47+
other, and better evolve the Web platform.</p>
48+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-4"></div>
4049
<p><cite>Anssi:</cite> On behalf of the program committee, I would
4150
like to thank the speakers for the incredible work you've produced.
4251
Special thanks to our sponsor Futurice for making this workshop
43-
possible.</p>
52+
possible.</p>
53+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-5"></div>
4454
<p><cite>Anssi:</cite> Today's session is on opportunities and
4555
challenges of browser based ML. We have divided the rest in 3 other
4656
slots. 4 sessions in total. Sep 22, 23, 29, same time as today for
47-
remaining sessions.</p>
57+
remaining sessions.</p>
58+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-7"></div>
4859
<p><cite>Dom:</cite> What is W3C? I'm Dominique Hazaël-Massieux,
4960
part of the technical staff at W3C. You may have received emails
5061
from me… W3C stands for the World Wide Web Consortium. We are
@@ -63,15 +74,17 @@ <h2>Introduction to the workshop</h2>
6374
are new for our scope. A big motivation for this workshop is that,
6475
as we see communities working with ML, we understand opportunities
6576
that ML creates for the Web, as well as opportunities that the Web
66-
creates for ML. Welcome all to this W3C event!</p>
77+
creates for ML. Welcome all to this W3C event!</p>
78+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-8"></div>
6779
<p><cite>Anssi:</cite> Here is an illustration of how we work
6880
together. Coming here, we received an array of talks. We have had
6981
very productive discussions on GitHub on some of the issues raised
7082
by the talk. As an outcome of this workshop, we will have a list of
7183
proposals, some of which may be ready for standardization
7284
(agreement and scope). We'll probably identify areas that are good
7385
target for incubation. We may also identify areas that would be
74-
better served in other organizations, not W3C.</p>
86+
better served in other organizations, not W3C.</p>
87+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-9"></div>
7588
<p><cite>Anssi:</cite> Please do not hesitate to comment using
7689
whatever mechanism you are comfortable with. Practically speaking,
7790
you are muted by default in Zoom to avoid background noise. Please
@@ -80,12 +93,14 @@ <h2>Introduction to the workshop</h2>
8093
to Zoom, we have a Slack channel, you may want to follow that as
8194
well. We take live notes through a scribe (Francois does it today).
8295
Lastly, we are operating under the Code of Ethics and Professional
83-
Conduct (CEPC).</p>
96+
Conduct (CEPC).</p>
97+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-10"></div>
8498
<p><cite>Anssi:</cite> Let's get to the meat of the meeting and
8599
look at the opportunities and challenges. Plenty of discussion
86100
topics for today in 3 buckets.</p>
87101
<h2>Improving existing web platform capabilities</h2>
88102
<h3>WebGPU fitness for ML frameworks</h3>
103+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-13"></div>
89104
<p><a href="https://github.com/w3c/machine-learning-workshop/issues/66">Related Discussions on Github</a></p>
90105
<p><cite>Anssi:</cite> First topic is WebGPU fitness for ML
91106
frameworks. Extensively discussed on GitHub. Does the WebGPU API
@@ -165,6 +180,7 @@ <h3>WebGPU fitness for ML frameworks</h3>
165180
progress. "Try different things and see what sticks". I would like
166181
to move to next topic now. Thank you for the great discussion.</p>
167182
<h3>Support for Float16 in JS & WebAssembly environments</h3>
183+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-14"></div>
168184
<p><a href="https://github.com/w3c/machine-learning-workshop/issues/64">Related Discussions on Github</a></p>
169185
<p><cite>Anssi:</cite> Lack of support in these environments is an
170186
issue for quantized models. There has been discussion a couple of
@@ -228,6 +244,7 @@ <h3>Comments on chat</h3>
228244
doesn't exist on ARM / GPU?</p>
229245
</div>
230246
<h3>Memory Copies</h3>
247+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-15"></div>
231248
<p><a href="https://github.com/w3c/machine-learning-workshop/issues/93">Related Discussions on Github</a></p>
232249
<p><cite>Anssi:</cite> The issue is that ML apps within the browser
233250
may trigger too many memory copies compared to native apps. This
@@ -317,6 +334,7 @@ <h3>Comments on chat</h3>
317334
doesn't solve it all)</p>
318335
</div>
319336
<h3>Permission model for Machine Learning APIs</h3>
337+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-16"></div>
320338
<p><a href="https://github.com/w3c/machine-learning-workshop/issues/72">Related Discussions on Github</a></p>
321339
<p><cite>Anssi:</cite> How to design a forward-looking permission
322340
model for ML APIs? This is a large problem, and not restricted to
@@ -408,6 +426,7 @@ <h3>Comments on chat</h3>
408426
</div>
409427
<h2>Extending beyond the browser</h2>
410428
<h3>Applicability to non-browser JS environments</h3>
429+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-18"></div>
411430
<p><cite>Anssi:</cite> Pay attention to the application of the
412431
browser-targeted work to non-browser JS runtimes, such as
413432
Node.js.</p>
@@ -446,6 +465,7 @@ <h3>Applicability to non-browser JS environments</h3>
446465
<p><cite>Anssi:</cite> I believe we have a separate topic for this
447466
discussion. So get in touch with TC53.</p>
448467
<p>[Time check, oh no, too late!]</p>
468+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-19"></div>
449469
<p><cite>Anssi:</cite> This was the first time we tried this,
450470
thanks for the discussion. We'll reschedule pending topics to the
451471
other 3 sessions. Virtual applause to everyone! I hope to see you

‎minutes/20200922.html

+22-9
Original file line numberDiff line numberDiff line change
@@ -3,6 +3,13 @@
33
layout: subpage
44
base: ../
55
parent: proceeding.html
6+
hasPDF: true
7+
prevlink: 20200916.html
8+
prevtitle: "Minutes of 1st Live Session (September 16, 2020)"
9+
nextlink: 20200929.html
10+
nexttitle: "Minutes of 3rd Live Session (September 23, 2020)"
11+
uplink: ../proceeding.html
12+
uptitle: List of minutes
613
---
714
<h2>Minutes of 2nd Live Session (September 22 2020)</h2>
815
<address><strong>Scribe:</strong> Dominique Hazael-Massieux</address>
@@ -19,8 +26,8 @@ <h2>Minutes of 2nd Live Session (September 22 2020)</h2>
1926
Cintron, Benjamin Bogdanovic, Jun Weifu, Mirza, Ping Yu, Chris
2027
Needham, Mark Crawford, Zhenjie</p>
2128
<p><strong>Regrets:</strong> Sangwhan Moon, François Daoust,</p>
22-
<p><strong>Agenda:</strong>
23-
<a href="https://github.com/w3c/machine-learning-workshop/tree/master/agenda#september-22-2020-2pm-utc-%EF%B8%8F">https://github.com/w3c/machine-learning-workshop/tree/master/agenda#september-22-2020-2pm-utc-%EF%B8%8F</a></p>
29+
<p><strong><a href="https://github.com/w3c/machine-learning-workshop/tree/master/agenda#september-22-2020-2pm-utc-%EF%B8%8F">Agenda</a></strong> - <strong><a href="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf#page=20">Slides</a></strong></p>
30+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-21"></div>
2431
<p><cite>Anssi:</cite> Introducing live session #2. What will we
2532
cover today? Last Tuesday, we discussed opportunities and
2633
challenges, today focuses on Web Platform foundations, broken into
@@ -33,12 +40,14 @@ <h2>Minutes of 2nd Live Session (September 22 2020)</h2>
3340
Please help correct the minutes as needed. We're operating under
3441
the Code of Ethics and Professional Conduct.</p>
3542
<p><cite>Dom:</cite> please make sure to speak slowly to help the
36-
interpreter</p>
43+
interpreter</p>
44+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-23"></div>
3745
<p><cite>Anssi:</cite> the goal of oru session today is to
3846
understand how machine learning fits in the Web technology stack;
3947
we have identified a number of discussion topics for today - we've
4048
picked "only" 6 topics today to leave enough time based on our
41-
experience last week.</p>
49+
experience last week.</p>
50+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-24"></div>
4251
<p>The first half of the meeting will touch on ML model formats,
4352
protecting ML models, in browser training, and training across
4453
devices.</p>
@@ -48,7 +57,8 @@ <h2>Minutes of 2nd Live Session (September 22 2020)</h2>
4857
<h2>Considerations for creating and deploying models</h2>
4958
<h3><a href=
5059
"https://github.com/w3c/machine-learning-workshop/issues/74">
51-
ML model format #74</a></h3>
60+
ML model format #74</a></h3>
61+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-26"></div>
5262
<p><cite>Anssi:</cite> Many talks touched on this topic. The high
5363
level issue is that there is no standard format for packaging &
5464
shipping ML models, and ML model formats evolve rapidly. A proposal
@@ -151,7 +161,8 @@ <h3><a href=
151161
understanding of what everyone is doing.</p>
152162
<h3><a href=
153163
"https://github.com/w3c/machine-learning-workshop/issues/67">
154-
Protecting ML models #67</a></h3>
164+
Protecting ML models #67</a></h3>
165+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-27"></div>
155166
<p><cite>Anssi:</cite> Protecting ML models; some ML providers need
156167
to ensure their models cannot be extracted from a browser app. E.g.
157168
the model is built and tuned to solve their business specific
@@ -239,12 +250,13 @@ <h3><a href=
239250
indeed</p>
240251
<p><cite>Anssi:</cite> Let's look into this</p>
241252
<div class="chat"><h3>Comments on Chat</h3>
242-
<p><cite>Christine&nbsp;Runnegar: +1</cite> to Wendy re security considerations,
253+
<p><cite>Christine&nbsp;Runnegar:</cite> +1 to Wendy re security considerations,
243254
and ditto for privacy</p>
244255
</div>
245256
<h3><a href=
246257
"https://github.com/w3c/machine-learning-workshop/issues/82">
247-
In-browser training #82</a></h3>
258+
In-browser training #82</a></h3>
259+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-28"></div>
248260
<p><cite>Anssi:</cite> our current efforts are focused on inference
249261
rather than training. The proposal under consideration is to
250262
understand real-world usages (e.g. Teachable Machine) and target
@@ -331,7 +343,8 @@ <h3><a href=
331343
better in the future for this kind of project.</p>
332344
<h3><a href=
333345
"https://github.com/w3c/machine-learning-workshop/issues/83">
334-
Training across devices #83</a></h3>
346+
Training across devices #83</a></h3>
347+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-29"></div>
335348
<p><cite>Anssi:</cite> this one is looking a bit further in the
336349
future - training across devices. The summary of this issue: we
337350
want to understand the role of edge computing in the platform. We

‎minutes/20200923.html

+20-7
Original file line numberDiff line numberDiff line change
@@ -3,7 +3,15 @@
33
layout: subpage
44
base: ../
55
parent: proceeding.html
6+
hasPDF: true
7+
prevlink: 20200922.html
8+
prevtitle: "Minutes of 2nd Live Session (September 22 2020)"
9+
nextlink: 20200929.html
10+
nexttitle: "Minutes of 4th Live Session (September 29 2020)"
11+
uplink: ../proceeding.html
12+
uptitle: List of minutes
613
---
14+
<form id=form></form>
715
<h2>Minutes of 3rd Live Session (September 23 2020)</h2>
816

917
<address><strong>Scribe:</strong> Dominique Hazael-Massieux</address>
@@ -21,8 +29,8 @@ <h2>Minutes of 3rd Live Session (September 23 2020)</h2>
2129
Solodovnikov, Josh, 高文灵, &nbsp;Wolfgang Maass, Stephan Steglich,
2230
Oleksandr Paraska, [Ceweb.br], Jesh</p>
2331
<p><strong>Regrets:</strong> Ruoy Ran</p>
24-
<p><strong>Agenda:</strong>
25-
<a href="https://github.com/w3c/machine-learning-workshop/tree/master/agenda#september-23-2020-2pm-utc-%EF%B8%8F">https://github.com/w3c/machine-learning-workshop/tree/master/agenda#september-23-2020-2pm-utc-%EF%B8%8F</a></p>
32+
<p><strong><a href="https://github.com/w3c/machine-learning-workshop/tree/master/agenda#september-23-2020-2pm-utc-%EF%B8%8F">Agenda</a></strong> - <strong><a href="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf#page=34">Slides</a></strong> </p>
33+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-35"></div>
2634
<p><cite>Anssi:</cite> this 3rd live session is focused on
2735
developer's perspective - including authoring ML experiences,
2836
reusing ML models, technical solutions and gaps. The topics on
@@ -36,7 +44,8 @@ <h2>Minutes of 3rd Live Session (September 23 2020)</h2>
3644
<h2>Applying web design principles to ML</h2>
3745
<h3><a href=
3846
"https://github.com/w3c/machine-learning-workshop/issues/68">
39-
Progressive Enhancement / Graceful degradation #68</a></h3>
47+
Progressive Enhancement / Graceful degradation #68</a></h3>
48+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-40"></div>
4049
<p><cite>Anssi:</cite> The question is how to bring ML features as
4150
optional improvements without breaking Web Compat? Dom provided
4251
some historical context - how the Web platform has evolved to cater
@@ -217,7 +226,8 @@ <h3><a href=
217226
our github issue</p>
218227
<h3><a href=
219228
"https://github.com/w3c/machine-learning-workshop/issues/80">
220-
Conformance testing of ML APIs for the Web #80</a></h3>
229+
Conformance testing of ML APIs for the Web #80</a></h3>
230+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-41"></div>
221231
<p><cite>Anssi:</cite> robust conformance testing has been the
222232
cornerstone of an interoperable Web platform; how to scale that
223233
practice to the ML APIs? Chai brought the challenges of numerical
@@ -303,7 +313,8 @@ <h3><a href=
303313
<h2>Improving web developer ergonomics</h2>
304314
<h3><a href=
305315
"https://github.com/w3c/machine-learning-workshop/issues/73">
306-
JS Operator overloading for Machine Learning #73</a></h3>
316+
JS Operator overloading for Machine Learning #73</a></h3>
317+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-43"></div>
307318
<p><cite>Sangwhan:</cite> I'm on the W3C TAG; I brought this up in
308319
my talk as something needed for an ML API to make sense on the Web.
309320
Using function call chaining deviates from how equations are
@@ -343,7 +354,8 @@ <h3><a href=
343354
interest on this?</p>
344355
<h3><a href=
345356
"https://github.com/w3c/machine-learning-workshop/issues/63">
346-
WebGL garbage collection #63</a></h3>
357+
WebGL garbage collection #63</a></h3>
358+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-44"></div>
347359
<p><cite>Anssi:</cite> we heard in the GH issue is that WebGL GC
348360
affects multiple ML libraries through side-effects. The proposal is
349361
to identify improvements to alleviate GC issues, ensure
@@ -375,7 +387,8 @@ <h3><a href=
375387
used in a while); definitely worth considering at the minimum.</p>
376388
<h3><a href=
377389
"https://github.com/w3c/machine-learning-workshop/issues/102">
378-
Neural network-oriented graph database #102</a></h3>
390+
Neural network-oriented graph database #102</a></h3>
391+
<div role="region" data-fmt=pdf data-src="http://www.w3.org/2020/Talks/mlws/live-sessions.pdf" id="slide-45"></div>
379392
<p><cite>Anssi:</cite> Wenhe Li in his talk on Pipcook made a point
380393
on storing ML in Web contexts; DL models are essentially weighted
381394
graphs, which would be better stored in a NN oriented database,

0 commit comments

Comments
 (0)