Sustaining Fairness via Incremental Learning

Authors: Somnath Basu Roy Chowdhury, Snigdha Chaturvedi

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical evaluations show that Fa IRL is able to make fair decisions while achieving high performance on the target task, outperforming several baselines.
Researcher Affiliation Academia Somnath Basu Roy Chowdhury, Snigdha Chaturvedi University of North Carolina at Chapel Hill {somnath, snigdha}@cs.unc.edu
Pseudocode Yes Algorithm 1: Prototype Sampling
Open Source Code Yes Our implementation of Fa IRL is publicly available at https://github.com/brcsomnath/Fa IRL.
Open Datasets Yes Biased MNIST. We follow the setup of (Bahng et al. 2020) to generate a synthetic dataset using MNIST (Le Cun et al. 1998)... Biography classification. We re-purpose the BIOS dataset (De-Arteaga et al. 2019)...
Dataset Splits No The paper discusses training and test sets and an exemplar-based approach for incremental learning. However, it does not explicitly define a separate validation split for hyperparameter tuning or model selection.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running the experiments in the provided text. It mentions "training modern machine learning systems" which implies hardware usage but without specifics.
Software Dependencies No The paper mentions "Our implementation of Fa IRL is publicly available at https://github.com/brcsomnath/Fa IRL." which might contain software details, but the provided text does not explicitly list any software dependencies with specific version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup No The paper mentions that "Additional details of our experimental setup can be found in Appendix B." However, Appendix B is not included in the provided text. The main body only mentions "β is a hyperparameter" (Equation 4) without specifying its value or other common experimental setup details like learning rate, batch size, or number of epochs.