Structured Bayesian Pruning via Log-Normal Multiplicative Noise

TL;DR

This paper introduces a Bayesian approach to structured neural network pruning using log-normal noise, enabling automatic removal of neurons or channels for faster inference while maintaining accuracy.

Contribution

It proposes a novel Bayesian model with structured sparsity that considers network architecture, using a truncated log-uniform prior and a closed-form variational approximation.

Findings

Achieves significant acceleration on various deep neural networks.

Automatically removes neurons or channels based on SNR.

Easy to implement as a dropout-like layer.

Abstract

Dropout-based regularization methods can be regarded as injecting random noise with pre-defined magnitude to different parts of the neural network during training. It was recently shown that Bayesian dropout procedure not only improves generalization but also leads to extremely sparse neural architectures by automatically setting the individual noise magnitude per weight. However, this sparsity can hardly be used for acceleration since it is unstructured. In the paper, we propose a new Bayesian model that takes into account the computational structure of neural networks and provides structured sparsity, e.g. removes neurons and/or convolutional channels in CNNs. To do this we inject noise to the neurons outputs while keeping the weights unregularized. We establish the probabilistic model with a proper truncated log-uniform prior over the noise and truncated log-normal variational approximation that ensures that the KL-term in the evidence lower bound is computed in closed-form. The model leads to structured sparsity by removing elements with a low SNR from the computation graph and provides significant acceleration on a number of deep neural architectures. The model is easy to implement as it can be formulated as a separate dropout-like layer.