wsl-workshop.github.io/acml22.html at master · wsl-workshop/wsl-workshop.github.io · GitHub

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
<!DOCTYPE html>
<html lang="en-US">
  <head>
    <title>ACML2022 WSL Workshop</title>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <link rel="stylesheet" href="./style.css">
  </head>

<script type="text/javascript">
var myDate = "2022-12-12T"
var myStartTime = "09:30"
var myTimeZone = "+05:30"

var myStartTime = new Date(myDate + myStartTime + ":00.000" + myTimeZone);
var myRefTime = new Date(myDate + "00:00:00.000" + myTimeZone);

function getTimezone() {
  var offset = -(new Date()).getTimezoneOffset()/60;
  return ("UTC" + (offset >= 0 ? "+" : "") + offset);
}

function getLocalTimezone() {
  try {
    return Intl.DateTimeFormat().resolvedOptions().timeZone + " time (" + getTimezone() + ", your browser's time zone)";
  }
  catch(e) {
    return " (" + getTimezone() + ", i.e., your browser's time zone)";
  }
}

function displayTime(dt) {
  var hour = dt.getHours();
  var minute = dt.getMinutes();
  var temp = '' + ((hour < 10) ? '0' : '') + hour;
  temp += ((minute < 10) ? ':0' : ':') + minute;
  return temp;
}

function writeTimeRange(startHour, startMin, endHour, endMin) {
  var oneMin = 1000 * 60;
  var oneHour = oneMin * 60;
  var startTime = new Date(myRefTime.getTime() + startHour * oneHour + startMin * oneMin);
  var endTime = new Date(myRefTime.getTime() + endHour * oneHour + endMin * oneMin);
  document.write(displayTime(startTime));
  document.write(" -- ");
  document.write(displayTime(endTime));
  if (endTime.getDay() != myStartTime.getDay()) {
    document.write(" (next day)");
  }
  return;
}
</script>


 <body>
    <section class="page-header">
      <h1 class="project-name">ACML 2022 Workshop</h1>
      <h1 class="project-name">Weakly Supervised Learning</h1>
      <h2 class="project-tagline">
      <noscript>Dec 12 9:30 -- 12:00 UTC+5:30 (the official time zone)</noscript>
      <script type="text/javascript">
        document.write(myStartTime.toLocaleDateString('en-US', {month:'short', day:'numeric'}));
        document.write(" ");
        writeTimeRange(10, 00, 13, 10);
        document.write(" ");
        document.write(getLocalTimezone())
      </script>
      </h2>
    </section>

<section class="main-content">
    <center>[ <a href="#program">Program</a>,
              <a href="#topics">Topics</a>,
              <a href="#organizers">Organizers</a>,
              <a href="#previous-workshops">Previous Workshops</a> ]</center>

<p>Note that this workshop is a part of <a href="http://www.acml-conf.org/2022/">Asian Conference on Machine Learning 2022</a>, and thus <font color='FF0000'>in order to attend the workshop, you should first register for the conference</font>. Then, the link of the airmeet can be found at the virtual site of ACML 2022 (if you attend the online conference).</p>

<p>Since ACML 2022 uses the AirMeet to deliver the online conference, please check that you can successfully log in ACML 2022 AirMeet (after you register ACML 2022, AirMeet Events will send you an email containing the “enter event” link). <font color='FF0000'>The email from AirMeet might be blocked, so please double check this.</font> If your web browser has remembered your login cookie, you should be able to join ACML 2022 via this <a href="https://www.airmeet.com/e/15da2c50-6f2b-11ed-8f6f-570b60bab04e">link</a>.</p>

<h1 id="program">Program</h1>

    <h2>Schedule</h1>
        <p>The workshop will use <font color='FF0000'>
        <script>document.write(getLocalTimezone());</script>
        <noscript>UTC+5:30 (the official time zone)</noscript>
        </font> for scheduling, and it will be combined with invited keynote talks and contributed talks (including invited short talks).</p>

<table>
  <thead>
    <tr><th style="width:25ex">Time</th><th>Event</th></tr>
  </thead>
  <tbody>
    <tr><td><script>writeTimeRange(10, 00, 10, 05);</script><noscript>10:00 -- 10:05</noscript></td>
        <td><b>Opening Ceremony</b></td></tr>
    <tr><td></td><td><b>Host</b>: <a href="http://www.ms.k.u-tokyo.ac.jp/sugi/" target="_blank">Masashi Sugiyama (he/him)</a></td></tr>

    <tr><td><script>writeTimeRange(10, 05, 10, 45);</script><noscript>10:05 -- 10:45</noscript></td>
        <td><b>Invited Keynote Talk 1</b></td></tr>
    <tr><td></td><td><b>Title</b>: Recent Progress and Challenges in Few-shot Learning</td></tr>
    <tr><td></td><td><b>Speaker</b>: <a href="http://www.cs.toronto.edu/~eleni/" target="_blank">Eleni Triantafillou (she/her)</a></td></tr>

    <tr><td><script>writeTimeRange(10, 45, 10, 55);</script><noscript>10:45 -- 10:55</noscript></td>
        <td><b>Contributed talks (Part I)</b></td></tr>
    <tr><td></td><td><b>Title</b>: MissDAG: Causal Discovery in the Presence of Missing Data with Continuous Additive Noise Models<br>
        <b>Speaker</b>: <a href="https://scholar.google.com.au/citations?user=vSJna-IAAAAJ&hl=en&oi=ao" target="_blank"> Erdun Gao (he/him)</a></td></tr>

    <tr><td><script>writeTimeRange(10, 55, 11, 30);</script><noscript>10:55 -- 11:30</noscript></td>
        <td><b>Bio Breaks and Social</b></td></tr>

    <tr><td><script>writeTimeRange(11, 30, 12, 10);</script><noscript>11:30 -- 12:10</noscript></td>
        <td><b>Contributed talks (Part II)</b></td></tr>
    <tr><td></td><td><b>Invited Short Talk</b>: Robust Semi-Supervised Learning for Open Environments<br>
        <b>Speaker</b>: <a href="http://www.lamda.nju.edu.cn/guolz/" target="_blank"> Lan-Zhe Guo (he/him)</a></td></tr>
    <tr><td></td><td><b>Title</b>: Escaping Saddle Points for Effective Generalization on Class-Imbalanced Data<br>
        <b>Speaker</b>: <a href="https://rangwani-harsh.github.io/about/" target="_blank">Harsh Rangwani (he/him)</a></td></tr>
    <tr><td></td><td><b>Invited Short Talk</b>: Simplified Graph Learning for Inductive Short Text Classification<br>
        <b>Speaker</b>: <a href="https://cse.hkust.edu.hk/~ywangcy/" target="_blank">Yaqing Wang (she/her)</a></td></tr>
    <tr><td></td><td><b>Title</b>: A Boosting Algorithm for Positive-Unlabeled Learning<br>
        <b>Speaker</b>: <a href="https://itee.uq.edu.au/profile/5606/yawen-zhao" target="_blank">Yawen Zhao (she/her)</a></td></tr>

    <tr><td><script>writeTimeRange(12, 10, 12, 50);</script><noscript>12:10 -- 12:50</noscript></td>
        <td><b>Invited Keynote Talk 2</b></td></tr>
    <tr><td></td><td><b>Title</b>: Causal Representation Learning: Successes and Challenges</td></tr>
    <tr><td></td><td><b>Speaker</b>: <a href="https://www.andrew.cmu.edu/user/kunz1/index.html" target="_blank">Kun Zhang (he/him)</a></td></tr>

    <tr><td><script>writeTimeRange(12, 50, 13, 10);</script><noscript>12:50 -- 13:10</noscript></td>
        <td><b>Panel Discussion and Concluding Remark</b></td></tr>
    <tr><td></td><td><b>Host</b>: <a href="https://fengliu90.github.io/" target="_blank">Feng Liu (he/him)</a></td></tr>
  </tbody>
</table>

    <h2>Invited Keynote Speakers</h2>

        <p><a href="http://www.cs.toronto.edu/~eleni/" target="_blank">Dr Eleni Triantafillou (she/her)</a>, Google Brain</p>
        <p><a href="https://www.andrew.cmu.edu/user/kunz1/index.html" target="_blank">Prof Kun Zhang (he/him)</a>, CMU and MBZUAI</p>

    <h2>Invited Short-talk Speakers</h2>

        <p><a href="http://www.lamda.nju.edu.cn/guolz/" target="_blank"> Dr Lan-Zhe Guo (he/him)</a>, LAMDA Group, Nanjing University</p>
        <p><a href="https://cse.hkust.edu.hk/~ywangcy/" target="_blank">Dr Yaqing Wang (she/her)</a>, Baidu Research</p>


<h1 id="topics">Topics</h1>

    <h2>Overview</h2>
        <p>Machine learning should not be accessible only to those who can pay. Specifically, modern machine learning is migrating to the era of complex models (e.g., deep neural networks), which require a plethora of well-annotated data. Giant companies have enough money to collect well-annotated data. However, for startups or non-profit organizations, such data is barely acquirable due to the cost of labeling data or the intrinsic scarcity in the given domain. These practical issues motivate us to research and pay attention to <b>weakly supervised learning (WSL)</b>, since WSL does not require such a huge amount of annotated data. We define WSL as the collection of machine learning problem settings and algorithms that share the same goals as supervised learning but can only access to <b>less supervised information</b> than supervised learning. In this workshop, we discuss both theoretical and applied aspects of WSL.</p>

        <p>This workshop is a series of our previous workshops at ACML 2019, SDM 2020, ACML 2020, IJCAI 2021, and ACML 2021. Our particular technical emphasis at this workshop is incomplete supervision, inexact supervision, inaccurate supervision, cross-domain supervision, imperfect demonstration, and weak adversarial supervision. Meanwhile, this workshop will also focus on <b>WSL for Science and Social Good</b>, such as WSL for COVID-19, WSL for healthcare, WSL for climate change, and WSL for remote sensing, and new public WSL datasets regarding the scientific scenarios. </p>

    <h2>Topics of Interest</h2>
        <p>WSL workshop includes (but not limited to) the following topics:</p>
        <ul>
            <li>Algorithms and theories of <b>incomplete supervision</b>, e.g., semi-supervised learning, active learning and positive-unlabeled learning;</li>
            <li>Algorithms and theories of <b>inexact supervision</b>, e.g., multi-instance learning, complementary learning, and open-set learning;</li>
            <li>Algorithms and theories of <b>inaccurate supervision</b>, e.g., crowdsourced learning and label-noise learning;</li>
            <li>Algorithms and theories of <b>cross-domain supervision</b>, e.g., zero-/one-/few-shot learning, transferable learning and multi-task leaning;</li>
            <li>Algorithms and theories of <b>imperfect demonstration</b>, e.g., inverse reinforcement learning and imitation learning with non-expert demonstrations;</li>
            <li>Algorithms and theories of <b>adversarial</b> weakly-supervised learning, e.g., adversarial semi-supervised learning and adversarial contrastive learning and adversarial label-noisy learning; </li>
            <li>Broad applications of weakly supervised learning in the field of computer science, such as weakly supervised object detection <b>(computer vision)</b>, weakly supervised sequence modeling <b>(natural language processing)</b>, weakly supervised cross-media retrieval <b>(information retrieval)</b>.</li>
            <li><b>WSL for science and social good</b>, such as WSL for COVID-19, WSL for healthcare, WSL for climate change, and WSL for remote sensing, meanwhile, new public datasets regarding the above WSL research directions (new focus).</li>
        </ul>

    <h2>Further Descriptions</h2>
        <p>The focus of this workshop is six types of weak supervision: incomplete supervision, inexact supervision, inaccurate supervision, cross-domain supervision, imperfect demonstration, and weak adversarial supervision, which are briefly introduced below. </p>


        <li><b>Incomplete supervision</b> considers a subset of training data given with ground-truth labels while the other data remain unlabeled, such as semi-supervised learning and positive-unlabeled learning. </li>

        <li><b>Inexact supervision</b> considers the situation where some supervision information is given but not as exacted as desired, i.e., only coarse-grained labels are available. For example, if we are considering to classify every pixel of an image, rather than the image itself, then ImageNet becomes a benchmark with inexact supervision. Besides, multi-instance learning belongs to inexact supervision, where we do not exactly know which instance in the bag corresponds to the given ground-truth label. </li>

        <li><b>Inaccurate supervision</b> considers the situation where the supervision information is not always the ground-truth, such as label-noise learning.</li>

        <li><b>Cross-domain supervision</b> considers the situation where the supervision information is scarce or even non-existent in the current domain but can be possibly derived from other domains. Examples of cross-domain supervision appear in zero-/one-/few-shot learning, where external knowledge from other domains is usually used to overcome the problem of too few or even no supervision in the original domain. </li>

        <li><b>Imperfect demonstration</b> considers the situation for inverse reinforcement learning and imitation learning, where the agent learns with imperfect or non-expert demonstrations. For example, AlphaGo learns a policy from a sequence of states and actions (expert demonstration). Even if an expert player wins a game, it is not guaranteed that every action in the sequence is optimal.</li>

        <li><b>Weak adversarial supervision</b> cconsiders the situation where weak supervision meets adversarial robustness. Since machine learning models are increasingly deployed in real-world applications, their security attracts more and more attention from both academia and industry. Therefore, many robust learning algorithms aim to prevent various evasion attacks, e.g., adversarial attacks, privacy attacks, model stealing attacks, and so on. However, almost all those robust algorithms (against evasion attacks) implicitly assume the strong supervision signals (no noisy labels in the training data), which hardly meets the requirements in practice. Therefore, when we develop evasion-robust algorithms, it is very practical/urgent to consider the supervision signals are imperfect.</li>

        <p>Meanwhile, this workshop will <b>continue discussing broad applications of weakly supervised learning in the field of computer science</b>, such as weakly supervised object detection (computer vision), weakly supervised sequence modeling (natural language processing), weakly supervised cross-media retrieval (information retrieval).</p>

        <p>More importantly, this workshop will focus on <b>how our society benefits from WSL methodologies, i.e., WSL for science and social good</b>. In many scientific scenarios, we will face 1) the data-shortage issue, e.g., medical records at the very beginning of the appearance of COVID-19, and 2) inaccurate observations, e.g., noisy labels of medical images, and 3) non-stationary data, e.g., meteorology records. Thus, how to learn from imperfect information is critical in science and should be noticed in the field of WSL. Studying these scientific scenarios will also boost the development of real-world datasets regarding the WSL methodologies, making researchers propose more algorithms that can make contributions to natural science. </p>


<h1 id="organizers">Organizers</h1>

    <h2>Program Co-chairs</h1>
        <p><a href="https://fengliu90.github.io/" target="_blank">Feng Liu</a>, The University of Melbourne, Australia.</p>
        <p><a href="https://zjfheart.github.io/" target="_blank">Jingfeng Zhang</a>, RIKEN, Japan.</p>
        <p><a href="https://scholar.google.co.jp/citations?user=KQUQlG4AAAAJ&hl=en" target="_blank">Nan Lu</a>, The University of Tokyo, Japan.</p>
        <p><a href="https://lfeng-ntu.github.io/" target="_blank">Lei Feng</a>, Chongqing University, China.</p>
        <p><a href="https://tongliang-liu.github.io/" target="_blank">Tongliang Liu</a>, The University of Sydney, Australia.</p>
        <p><a href="https://bhanml.github.io/" target="_blank">Bo Han</a>, Hong Kong Baptist University, Hong Kong SAR, China.</p>
        <p><a href="https://niug1984.github.io/" target="_blank">Gang Niu</a>, RIKEN, Japan.</p>
        <p><a href="http://www.ms.k.u-tokyo.ac.jp/sugi/" target="_blank">Masashi Sugiyama</a>, RIKEN / University of Tokyo, Japan.</p>

    <h2>Advisory Board <span style="font-size: 75%">(alphabetical order by last name)</span></h1>
        <p><a href="https://gcatnjust.github.io/ChenGong/index.html" target="_blank">Chen Gong</a>, Nanjing University of Science and Technology, China.</p>
        <p><a href="https://mingming-gong.github.io/" target="_blank">Mingming Gong</a>, The University of Melbourne, Australia.</p>
        <p><a href="https://cs.nju.edu.cn/liyf/index.htm" target="_blank">Yu-Feng Li</a>, Nanjing University, China.</p>
        <p><a href="http://www.yliuu.com/" target="_blank">Yang Liu</a>, University of California, Santa Cruz, US.</p>
        <p><a href="https://www.a-star.edu.sg/cfar/about-cfar/management/prof-ivor-tsang" target="_blank">Ivor W. Tsang</a>, A*STAR Centre for Frontier AI Research (CFAR), Singapore.</p>
        <p><a href="https://miaoxu-ml.github.io/index.html" target="_blank">Miao Xu</a>, The University of Queensland, Australia.</p>
        <p><a href="http://www.cse.ust.hk/~qyaoaa/" target="_blank">Quanming Yao</a>, Tsinghua University / 4Paradigm Inc., China.</p>
        <p><a href="http://palm.seu.edu.cn/zhangml/" target="_blank">Min-Ling Zhang</a>, Southeast University, China.</p>

<h1 id="previous-workshops">Previous Workshops</h1>
    <p><a href="https://wsl-workshop.github.io/acml21.html" target="_blank">ACML2021 WSL Workshop</a>, Online.</p>
    <p><a href="https://wsl-workshop.github.io/ijcai21.html" target="_blank">IJCAI2021 WSRL Workshop</a>, Online.</p>
    <p><a href="https://wsl-workshop.github.io/acml20.html" target="_blank">ACML2020 WSRL Workshop</a>, Online.</p>
    <p><a href="https://wsl-workshop.github.io/sdm20.html" target="_blank">SDM2020 WSUL Workshop</a>, Ohio, United States.</p>
    <p><a href="https://wsl-workshop.github.io/acml19.html" target="_blank">ACML2019 WSL Workshop</a>, Nagoya, Japan.</p>

<footer class="site-footer">
    <span class="site-footer-credits">This page was generated by <a href="https://pages.github.com/">GitHub Pages</a>.</span>
</footer>
</section>
</body></html>