Learning Debiased Classifier with Biased Committee. (arXiv:2206.10843v1 [cs.LG])

Neural networks are prone to be biased towards spurious correlations between
classes and latent attributes exhibited in a major portion of training data,
which ruins their generalization capability. This paper proposes a new method
for training debiased classifiers with no spurious attribute label. The key
idea of the method is to employ a committee of classifiers as an auxiliary
module that identifies bias-conflicting data, i.e., data without spurious
correlations, and assigns large weights to them when training the main
classifier. …

arxiv learning lg