Adversarially Robust PAC Learnability of Real-Valued Functions

Authors: Idan Attias, Steve Hanneke

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical From the theoretical standpoint, there has been a lot of effort to provide provable guarantees of such methods (e.g., Feige et al. (2015); Schmidt et al. (2018); Khim & Loh (2018); Yin et al. (2019); Cullina et al. (2018); Attias et al. (2019; 2022); Montasser et al. (2021b; 2020a;b; 2021a); Ashtiani et al. (2020); Dan et al. (2020); Awasthi et al. (2020; 2021b;a; 2022a;b; 2023); Bhattacharjee et al. (2021); Xing et al. (2021); Mao et al. (2023)), which is the focus of this work. Our main technique is based on a construction of an adversarially robust sample compression scheme of a size determined by the fat-shattering dimension.
Researcher Affiliation Academia 1Department of Computer Science, Ben-Gurion University, Israel 2Department of Computer Science, Purdue University, USA.
Pseudocode Yes Algorithm 1 Improper Robust Regressor with High-Vote, Algorithm 2 Improper Robust Regressor, Algorithm 3 Improper Robust (η, β)-Regressor, Algorithm 4 Modified Multiplicative Weights, Algorithm 5 Modified Med Boost, Algorithm 6 Sparsify, Algorithm 7
Open Source Code No The paper does not contain any statement about making its source code available or provide a link to a code repository.
Open Datasets No The paper is theoretical and does not mention specific public datasets, nor does it provide links or citations for data access.
Dataset Splits No The paper is theoretical and does not describe experimental data splits (training, validation, or test sets).
Hardware Specification No The paper is theoretical and does not describe any specific hardware used for experiments.
Software Dependencies No The paper is theoretical and does not mention specific software dependencies with version numbers.
Experiment Setup No The paper is theoretical and does not describe concrete experimental setup details, hyperparameters, or training configurations.