Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples

TL;DR

This paper introduces methods to improve neural network training by emphasizing high-variance samples, leading to more accurate models across diverse architectures and training techniques.

Contribution

It proposes lightweight, uncertainty-based sample weighting strategies that enhance training effectiveness beyond existing methods.

Findings

Improved accuracy across six datasets and multiple architectures.

Additional gains when combined with popular training techniques.

Reliable enhancement of model performance through variance-based sampling.

Abstract

Self-paced learning and hard example mining re-weight training instances to improve learning accuracy. This paper presents two improved alternatives based on lightweight estimates of sample uncertainty in stochastic gradient descent (SGD): the variance in predicted probability of the correct class across iterations of mini-batch SGD, and the proximity of the correct class probability to the decision threshold. Extensive experimental results on six datasets show that our methods reliably improve accuracy in various network architectures, including additional gains on top of other popular training techniques, such as residual learning, momentum, ADAM, batch normalization, dropout, and distillation.