Uncertainty Estimation via Stochastic Batch Normalization
This work addresses uncertainty estimation for deep learning practitioners, offering an incremental improvement over existing batch normalization methods.
The authors tackled the problem of computationally inefficient uncertainty estimation in neural networks by proposing Stochastic Batch Normalization, which provides a scalable approximation that achieves competitive performance on MNIST and CIFAR-10 datasets with popular architectures like VGG and ResNets.
In this work, we investigate Batch Normalization technique and propose its probabilistic interpretation. We propose a probabilistic model and show that Batch Normalization maximazes the lower bound of its marginalized log-likelihood. Then, according to the new probabilistic model, we design an algorithm which acts consistently during train and test. However, inference becomes computationally inefficient. To reduce memory and computational cost, we propose Stochastic Batch Normalization -- an efficient approximation of proper inference procedure. This method provides us with a scalable uncertainty estimation technique. We demonstrate the performance of Stochastic Batch Normalization on popular architectures (including deep convolutional architectures: VGG-like and ResNets) for MNIST and CIFAR-10 datasets.