Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples
This work addresses the problem of enhancing training efficiency for neural networks, though it is incremental as it builds on existing re-weighting techniques.
The paper tackled improving neural network training accuracy by re-weighting samples based on uncertainty estimates, achieving reliable accuracy gains across six datasets and various architectures.
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.