DAIS: Automatic Channel Pruning via Differentiable Annealing Indicator Search

TL;DR

DAIS is an automated channel pruning method that uses differentiable annealing search to optimize network structure, achieving better performance and efficiency than previous methods on standard datasets.

Contribution

Introduces DAIS, a neural architecture search-based approach for automatic channel pruning with annealing, reducing reliance on manual rules and improving pruning effectiveness.

Findings

DAIS outperforms state-of-the-art pruning methods on CIFAR-10, CIFAR-100, and ImageNet.

The annealing procedure effectively guides the model towards binarized channel indicators.

Regularizations based on structural knowledge improve pruning sparsity and model performance.

Abstract

The convolutional neural network has achieved great success in fulfilling computer vision tasks despite large computation overhead against efficient deployment. Structured (channel) pruning is usually applied to reduce the model redundancy while preserving the network structure, such that the pruned network can be easily deployed in practice. However, existing structured pruning methods require hand-crafted rules which may lead to tremendous pruning space. In this paper, we introduce Differentiable Annealing Indicator Search (DAIS) that leverages the strength of neural architecture search in the channel pruning and automatically searches for the effective pruned model with given constraints on computation overhead. Specifically, DAIS relaxes the binarized channel indicators to be continuous and then jointly learns both indicators and model parameters via bi-level optimization. To bridge the non-negligible discrepancy between the continuous model and the target binarized model, DAIS proposes an annealing-based procedure to steer the indicator convergence towards binarized states. Moreover, DAIS designs various regularizations based on a priori structural knowledge to control the pruning sparsity and to improve model performance. Experimental results show that DAIS outperforms state-of-the-art pruning methods on CIFAR-10, CIFAR-100, and ImageNet.