Learning from Failure: Training Debiased Classifier from Biased Classifier
This addresses the issue of biased AI models for users needing fair and robust predictions, offering a generic, cheaper alternative to supervised debiasing methods, though it builds incrementally on prior work by leveraging training dynamics.
The paper tackles the problem of neural networks learning biased predictions due to spurious correlations in datasets, proposing a failure-based debiasing scheme that trains a pair of networks to amplify and counteract bias, resulting in significant improvements across synthetic and real-world datasets, occasionally outperforming methods requiring explicit supervision.
Neural networks often learn to make predictions that overly rely on spurious correlation existing in the dataset, which causes the model to be biased. While previous work tackles this issue by using explicit labeling on the spuriously correlated attributes or presuming a particular bias type, we instead utilize a cheaper, yet generic form of human knowledge, which can be widely applicable to various types of bias. We first observe that neural networks learn to rely on the spurious correlation only when it is "easier" to learn than the desired knowledge, and such reliance is most prominent during the early phase of training. Based on the observations, we propose a failure-based debiasing scheme by training a pair of neural networks simultaneously. Our main idea is twofold; (a) we intentionally train the first network to be biased by repeatedly amplifying its "prejudice", and (b) we debias the training of the second network by focusing on samples that go against the prejudice of the biased network in (a). Extensive experiments demonstrate that our method significantly improves the training of the network against various types of biases in both synthetic and real-world datasets. Surprisingly, our framework even occasionally outperforms the debiasing methods requiring explicit supervision of the spuriously correlated attributes.