Towards Optimal Filter Pruning with Balanced Performance and Pruning Speed

TL;DR

This paper introduces a balanced filter pruning method that achieves near-optimal performance and fast pruning speed by layer-wise pruning based on loss variation, applicable to various architectures with minimal retraining.

Contribution

The proposed method enables layer-wise pruning without iterative retraining, efficiently finds pruning thresholds, and incorporates layer group pruning for networks with short connections.

Findings

Outperforms many state-of-the-art pruning methods.

Achieves a good balance between model performance and pruning speed.

Applicable to common neural network architectures.

Abstract

Filter pruning has drawn more attention since resource constrained platform requires more compact model for deployment. However, current pruning methods suffer either from the inferior performance of one-shot methods, or the expensive time cost of iterative training methods. In this paper, we propose a balanced filter pruning method for both performance and pruning speed. Based on the filter importance criteria, our method is able to prune a layer with approximate layer-wise optimal pruning rate at preset loss variation. The network is pruned in the layer-wise way without the time consuming prune-retrain iteration. If a pre-defined pruning rate for the entire network is given, we also introduce a method to find the corresponding loss variation threshold with fast converging speed. Moreover, we propose the layer group pruning and channel selection mechanism for channel alignment in network with short connections. The proposed pruning method is widely applicable to common architectures and does not involve any additional training except the final fine-tuning. Comprehensive experiments show that our method outperforms many state-of-the-art approaches.