Timezone: »

 
Poster
A Near-Optimal Algorithm for Debiasing Trained Machine Learning Models
Ibrahim Alabdulmohsin · Mario Lucic

Thu Dec 09 08:30 AM -- 10:00 AM (PST) @
in Poster Session 6 »

We present a scalable post-processing algorithm for debiasing trained models, including deep neural networks (DNNs), which we prove to be near-optimal by bounding its excess Bayes risk. We empirically validate its advantages on standard benchmark datasets across both classical algorithms as well as modern DNN architectures and demonstrate that it outperforms previous post-processing methods while performing on par with in-processing. In addition, we show that the proposed algorithm is particularly effective for models trained at scale where post-processing is a natural and practical choice.

Keywords: Deep Learning · Machine Learning · Fairness

Author Information

Ibrahim Alabdulmohsin (Google Research)
Mario Lucic (Google Brain)

More from the Same Authors