Fair Representations by Compression

Authors: Xavier Gitiaux, Huzefa Rangwala11506-11515

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results show that the proposed method, FBC, achieves state-of-the-art accuracyfairness trade-off. The objective of this experimental section is to demonstrate that Fairness by Binary Compression FBC can achieve state-of-the art performance compared to four benchmarks in fair representations learning: β-VAE, Adv, MMD and VFAE.
Researcher Affiliation Academia Xavier Gitiaux, Huzefa Rangwala George Mason University xgitiaux@gmu.edu, rangwala@gmu.edu
Pseudocode No The paper describes the method's steps in paragraph form, but it does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statements or links indicating that the source code for the described methodology (FBC) is publicly available.
Open Datasets Yes First, we apply our experimental protocol to a synthetic dataset DSprites Unfair, 2 that contains 64 by 64 black and white images of various shapes (heart, square, circle). (Footnote 2: https://github.com/deepmind/dsprites-dataset/); Then, we extend our experimental protocol to three benchmark datasets in fair machine learning: Adults, Compas and Heritage. The Adults dataset 3 contains 49K individuals... (Footnote 3: https://archive.ics.uci.edu/ml/datasets/adult); The Compas data 4 contains 7K individuals... (Footnote 4: https://github.com/propublica/compas-analysis/); The Health Heritage dataset 5 contains 220K individuals... (Footnote 5: https://foreverdata.org/1015/index.html)
Dataset Splits No Datasets are split into a training set used to trained the encoderdecoder architecture; two test sets, one to train both task and auditing networks on samples not seen by the encoderdecoder; one to evaluate their respective performances. The paper does not explicitly mention a 'validation' dataset split or the proportions of the splits.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments.
Software Dependencies No The paper mentions 'Pytorch' and 'Pixel CNN' but does not provide specific version numbers for these or any other software dependencies, which are necessary for reproducibility.
Experiment Setup No The paper mentions hyperparameters like 'σ' for soft-binarization and 'β' for rate-distortion trade-off, and states that 'we vary the value of the parameter β', but it does not provide specific numerical values for these or other critical experimental setup details such as learning rate, batch size, or optimizer settings.