Workshop

WeaSuL 2021


Workshop on Weakly Supervised Learning

ICLR 2021 Workshop
May 7 2021
virtual


Schedule

Day: May 7 2021
Location: Virtual (ICLR platform)

  • All talks will be streamed during the workshop on the conference webpage.
  • All talks have a live Q&A session and we have a panel discussion in the end. You can ask questions via Rocket.Chat or join the Zoom session directly. The links are also on the conference webpage.
  • The poster sessions and the post workshop hang-out are held on Gather.Town. Links below in the schedule.
  • Tweet with us @wea_su #WeaSuL2021
  • For those new to the topic, we collected a list of papers to get an overview over the field.
PDT EDT CEST BJT Event
07:00 10:00 16:00 22:00 Introduction and Opening Remarks
07:10 10:10 16:10 22:10 Invited Talk
Dan Roth (University of Pennsylvania)
08:25 11:25 17:25 23:25 Invited Talk
Marine Carpuat (University of Maryland)
09:25 12:25 18:25 00:25 Contributed Talk
Dependency Structure Misspecification in Multi-Source Weak Supervision Models
09:50 12:50 18:50 00:50 Poster Spotlights 1
10:15 13:15 19:15 01:15 Virtual Poster Session 1
        Link to Gather.Town Poster Room 1
        AutoTriggER: Named Entity Recognition with Auxiliary Trigger Extraction
        Handling Long-Tail Queries with Slice-Aware Conversational Systems
        Tabular Data Modeling via Contextual Embeddings
        TADPOLE: Task ADapted Pre-training via anOmaLy dEtection
        Active WeaSuL: Improving Weak Supervision with Active Learning
        Transformer Language Models as Universal Computation Engines
11:15 14:15 20:15 02:15 Welcome Back
11:20 14:20 20:20 02:20 Invited Talk
Heng Ji (University of Illinois)
12:05 15:05 21:05 03:05 Contributed Talk
Weakly Supervised Multi-task Learning for Concept-based Explainability
12:30 15:30 21:30 03:30 Contributed Talk
Better adaptation to distribution shifts with Robust Pseudo-Labeling
12:55 15:55 21:55 03:55 Poster Spotlights 2
13:20 16:20 22:20 04:20 Virtual Poster Session 2
        Link to Gather.Town Poster Room 2
        Using system context information to complement weakly labeled data
        CIGMO: Learning categorical invariant deep generative models from grouped data
        Pre-Training by Completing Points Cloud
        Weakly-Supervised Group Disentanglement using Total Correlation
        Is Disentanglement all you need? Comparing Concept-based & Disentanglement Approaches
        Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks
        Weakly Supervision Multi-Task Learning for Concept-Based Explainability
14:20 17:20 23:20 05:20 Welcome Back
14:25 17:25 23:25 05:25 Invited Talk
Lu Jiang (Google Research)
15:25 18:25 00:25 06:25 Invited Talk
Paroma Varma (Snorkel AI)
16:10 19:10 01:10 07:10 Virtual Panel Discussion
17:10 20:10 02:10 08:10 Conclusion
17:25 20:25 02:25 08:25 Post Workshop Hangout
        Link to Gather.Town Hangout