Multi-group Agnostic PAC Learnability

Authors: Guy N Rothblum, Gal Yona

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical Our main technical contributions answer both questions: 1. We prove a partial information-theoretic characterization of the compatible loss functions. 2. For any such loss function that also satisfies a natural uniform convergence property, we show an algorithm that, for any specified finite collection G and finite hypothesis class H, learns a multi-group agnostic predictor from labeled data.
Researcher Affiliation Academia 1Weizmann Institute of Science, Rehovot, Israel.
Pseudocode Yes Algorithm 1 Multi Group L,f(ϵ, δ, γ, H, G) and Algorithm 2 AL,f,k g,h,α (multi-sample Sample-Access OI distinguisher)
Open Source Code No The paper does not mention any open-source code for the methodology described.
Open Datasets No The paper does not describe the use of any specific dataset for training, as it focuses on theoretical analysis rather than empirical evaluation.
Dataset Splits No The paper is theoretical and does not describe empirical experiments, therefore no training/validation/test dataset splits are provided.
Hardware Specification No The paper is theoretical and does not conduct experiments, so no hardware specifications are mentioned.
Software Dependencies No The paper is theoretical and does not describe implementation details or empirical experiments, therefore no software dependencies with specific version numbers are provided.
Experiment Setup No The paper is theoretical and does not describe empirical experiments, thus no experimental setup details like hyperparameters or training settings are provided.