Too Relaxed to Be Fair

Authors: Michael Lohaus, Michael Perrot, Ulrike Von Luxburg

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we empirically evaluate Search Fair by comparing it to 5 baselines on 6 real-world datasets.
Researcher Affiliation Academia 1 University of Tubingen, Germany 2 Max Planck Institute for Intelligent Systems, Tubingen, Germany 3 Univ Lyon, UJM-Saintsimilarity-based classifiers to exhibit sufficient conditions Etienne, CNRS, IOGS, Lab HC UMR 5516, F-42023, SAINTthat guarantee the existence of a fair and accurate classifier. ETIENNE, France
Pseudocode Yes We call this procedure Search Fair and summarize it in Algorithm 1 in the supplementary.
Open Source Code Yes The code is freely available online: github.com/mlohaus/Search Fair.
Open Datasets Yes We consider different datasets: Celeb A (Liu et al., 2015), Adult (Kohavi & Becker, 1996), Dutch (Zliobaite et al., 2011), COMPAS (Larson et al., 2016), Communities and Crime (Redmond & Baveja, 2002), and German Credit (Dua & Graff, 2017).
Dataset Splits Yes We use 5-fold cross validation to choose other hyperparameters.
Hardware Specification No The paper does not provide specific hardware details such as GPU or CPU models used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python 3.x, TensorFlow 2.x).
Experiment Setup Yes We use 5-fold cross validation to choose other hyperparameters. [...] For all remaining methods we need to choose the regularization parameter β and the width of the rbf kernel. These values are respectively chosen in the sets 10−6, 10−5, 10−4, 10−3, 10−2 and {10log d−1 , 10log d , d, 10log d+1 , 10log d+2 }, with d the number of features.