Fair Mixup: Fairness via Interpolation

Authors: Ching-Yao Chuang, Youssef Mroueh

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

Reproducibility Variable Result LLM Response
Research Type Experimental We analyze fair mixup and empirically show that it ensures a better generalization for both accuracy and fairness measurement in tabular, vision, and language benchmarks.
Researcher Affiliation Collaboration Ching-Yao Chuang CSAIL, MIT cychuang@mit.edu Youssef Mroueh IBM Research AI mroueh@us.ibm.com
Pseudocode No The paper does not contain any pseudocode or algorithm blocks.
Open Source Code Yes The code is available at https://github.com/chingyaoc/fair-mixup.
Open Datasets Yes UCI Adult dataset (Dua & Graff, 2017) ... Celeb A face attributes dataset (Liu et al.)... Jigsaw toxic comment dataset (Jigsaw, 2018).
Dataset Splits Yes the dataset is randomly randomly split into a training, validation, and testing set with partition 60%, 20%, and 20%, respectively.
Hardware Specification No The paper does not specify the hardware (e.g., GPU/CPU models) used for running the experiments.
Software Dependencies No The paper mentions the Adam optimizer and BERT embeddings but does not provide specific version numbers for software dependencies like PyTorch, TensorFlow, or specific library versions.
Experiment Setup Yes The models are two-layer Re LU networks with hidden size 200. We only evaluate input mixup for Adult dataset as the network is not deep enough to produce meaningful latent representations. The models are optimized with Adam optimizer (Kingma & Ba, 2014) with learning rate 1e-3. We retrain each model 10 times and report the mean accuracy and fairness measurement.