Regularizing by the Variance of the Activations' Sample-Variances

Abstract

Normalization techniques play an important role in supporting efficient and often more effective training of deep neural networks. While conventional methods explicitly normalize the activations, we suggest to add a loss term instead. This new loss term encourages the variance of the activations to be stable and not vary from one random mini-batch to the next. As we prove, this encourages the activations to be distributed around a few distinct modes. We also show that if the inputs are from a mixture of two Gaussians, the new loss would either join the two together, or separate between them optimally in the LDA sense, depending on the prior probabilities. Finally, we are able to link the new regularization term to the batchnorm method, which provides it with a regularization perspective. Our experiments demonstrate an improvement in accuracy over the batchnorm technique for both CNNs and fully connected networks.

References in corpus (8)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Instance Normalization: The Missing Ingredient for Fast Stylization
• Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1
• Improved Techniques for Training GANs
• Learning multiple visual domains with residual adapters
• Group Sparse Regularization for Deep Neural Networks
• Learning Structured Sparsity in Deep Neural Networks
• Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models

Cited by in corpus (3)

• Normalization Techniques in Training DNNs: Methodology, Analysis and Application
• Four Things Everyone Should Know to Improve Batch Normalization
• Why Not to Use Zero Imputation? Correcting Sparsity Bias in Training Neural Networks