Individually Fair Learning with One-Sided Feedback
Authors: Yahav Bechavod, Aaron Roth
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | Our contributions lie on a conceptual as well as a technical level. We first present a novel auditing scheme... We then present our online learning framework... Our main technical contributions are given in Sections 3, 4. We first present an efficient reduction to the contextual combinatorial semi-bandit setting... We then establish an equivalence... Finally, we present an oracle-efficient algorithm for our setting and analyze the resulting rates for accuracy and fairness (Section 5). ... The guarantees of Theorem 5.3 can be interpreted as follows: accuracy-wise, the resulting algorithm is competitive with the performance of the most accurate policy that is fair (i.e. in Qα ϵ,γ). Fairness-wise, the number of rounds in which there exist (one or more) fairness violations, is sub-linear. |
| Researcher Affiliation | Academia | 1Hebrew University and University of Pennsylvania 2University of Pennsylvania. Correspondence to: Yahav Bechavod <yahav@seas.upenn.edu>, Aaron Roth <aaroth@cis.upenn.edu>. |
| Pseudocode | Yes | Algorithm 1 Individually Fair Online Learning with One Sided Feedback; Algorithm 2 Contextual Combinatorial Semi-Bandit; Algorithm 3 Reduction to Contextual Combinatorial Semi Bandit; Algorithm 4 Utilization of Context-Semi-Bandit-FTPL; Algorithm 5 Context-Semi-Bandit-FTPL-With-Resampling |
| Open Source Code | No | The paper is theoretical and focuses on algorithm design and guarantees, not on providing an implementation or releasing source code. |
| Open Datasets | No | The paper is theoretical, defining a learning problem and proposing algorithms, but does not conduct empirical evaluations on specific datasets. It mentions a 'loan approval setting' as a running example but does not specify any actual dataset for training. |
| Dataset Splits | No | The paper is theoretical and does not describe empirical experiments with data splits. |
| Hardware Specification | No | The paper is theoretical and describes algorithms and their properties, not their implementation on specific hardware. Therefore, no hardware specifications are mentioned. |
| Software Dependencies | No | The paper is theoretical and describes algorithms and their properties, not specific software implementations with version numbers. Therefore, no software dependencies with version numbers are mentioned. |
| Experiment Setup | No | The paper is theoretical, focusing on algorithmic design and mathematical guarantees, and therefore does not detail empirical experimental setups like hyperparameters or training configurations. |