Web: http://arxiv.org/abs/2209.06235

Sept. 15, 2022, 1:12 a.m. | Yann Dubois, Tatsunori Hashimoto, Stefano Ermon, Percy Liang

stat.ML updates on arXiv.org arxiv.org

Despite the empirical successes of self-supervised learning (SSL) methods, it
is unclear what characteristics of their representations lead to high
downstream accuracies. In this work, we characterize properties that SSL
representations should ideally satisfy. Specifically, we prove necessary and
sufficient conditions such that for any task invariant to given data
augmentations, desired probes (e.g., linear or MLP) trained on that
representation attain perfect accuracy. These requirements lead to a unifying
conceptual framework for improving existing SSL methods and deriving new …

arxiv self-supervised learning supervised learning

Research Scientists

@ ODU Research Foundation | Norfolk, Virginia

Embedded Systems Engineer (Robotics)

@ Neo Cybernetica | Bedford, New Hampshire

2023 Luis J. Alvarez and Admiral Grace M. Hopper Postdoc Fellowship in Computing Sciences

@ Lawrence Berkeley National Lab | San Francisco, CA

Senior Manager Data Scientist

@ NAV | Remote, US

Senior AI Research Scientist

@ Earth Species Project | Remote anywhere

Research Fellow- Center for Security and Emerging Technology (Multiple Opportunities)

@ University of California Davis | Washington, DC

Staff Fellow - Data Scientist

@ U.S. FDA/Center for Devices and Radiological Health | Silver Spring, Maryland

Staff Fellow - Senior Data Engineer

@ U.S. FDA/Center for Devices and Radiological Health | Silver Spring, Maryland

Senior Research Engineer, Applied Language

@ DeepMind | Mountain View, California, US

Machine Learning Engineer

@ Bluevine | Austin, TX

Lead Manager - Analytics & Data Science

@ Tide | India(Remote)

Machine Learning Engineer

@ Gtmhub | Indore, Madhya Pradesh, India