AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Fundamental Limits and Tradeoffs in Invariant Representation Learning

Han Zhao, Chen Dan, Bryon Aragam, Tommi Jaakkola, Geoffrey Gordon, Pradeep Ravikumar

NeurIPS 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use an MLP representation instead of LogitUCI Adult · UCI AdultI(Y;Z): 0.170 -> 0.216Improved
Use an MLP representation instead of LogitUCI Adult · UCI AdultI(A;Z): 0.103 -> 0.092Improved
Use an MLP representation instead of LinearUCI Adult · UCI AdultI(Y;Z): 0.171 -> 0.216Improved
Use an MLP representation instead of LinearUCI Adult · UCI AdultI(A;Z): 0.109 -> 0.092Improved
Use Adv instead of MLPUCI Adult · UCI AdultI(Y;Z): 0.216 -> 0.202Hurt
Use Adv instead of MLPUCI Adult · UCI AdultI(A;Z): 0.092 -> 0.068Improved
Use MLP instead of OLSLaw School · Law SchoolVar E[Y|Z]: 0.014 -> 0.022Improved
Use MLP instead of OLSLaw School · Law SchoolVar E[A|Z]: 0.008 -> 0.012Hurt
Use Adv instead of OLSLaw School · Law SchoolVar E[Y|Z]: 0.014 -> 0.018Improved
Use Adv instead of OLSLaw School · Law SchoolVar E[A|Z]: 0.008 -> 0.010Hurt
Randomize between existing representationsInvariant representation learning · Classification and regression information planesAchievable accuracy-invariance tradeoffsImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy