On Optimal Learning Under Targeted Data Poisoning

Authors: Steve Hanneke, Amin Karbasi, Mohammad Mahmoody, Idan Mehalel, Shay Moran

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this work we aim to characterize the smallest achievable error. We fully achieve this in the realizable setting, proving that ε = Ω(VC(H) γ). We then also show a matching lower bound.
Researcher Affiliation Collaboration Steve Hanneke Purdue University, USA; Amin Karbasi Yale University, USA and Google Research; Mohammad Mahmoody University of Virginia, USA; Idan Mehalel Technion, Israel; Shay Moran Technion, Israel and Google Research
Pseudocode Yes Figure 1: SPV A meta algorithm implementing a stable version of the input learning algorithm Lrn.
Open Source Code No The paper does not provide an explicit statement about releasing code or a link to a source code repository. The 'Checklist' sections regarding code are marked 'N/A'.
Open Datasets No The paper is theoretical and does not conduct empirical studies with specific datasets. The checklist explicitly states 'N/A' for data and experimental results.
Dataset Splits No The paper is theoretical and does not conduct empirical studies with specific datasets, therefore it does not mention training, validation, or test splits.
Hardware Specification No The paper is theoretical and does not report on empirical experiments, therefore no hardware specifications are mentioned. The 'Checklist' sections regarding compute resources are marked 'N/A'.
Software Dependencies No The paper is theoretical and does not report on empirical experiments, therefore no specific software dependencies with version numbers are listed. The 'Checklist' sections regarding compute resources are marked 'N/A'.
Experiment Setup No The paper is theoretical and focuses on mathematical proofs and algorithm design, not on the experimental setup, hyperparameters, or training details of empirical experiments. The 'Checklist' sections regarding training details are marked 'N/A'.