Learning to Prune Filters in Convolutional Neural Networks

TL;DR

This paper introduces a data-driven, learnable pruning method for CNNs that effectively reduces model size and computational cost while maintaining accuracy, using a novel reward-based training algorithm.

Contribution

It proposes a 'try-and-learn' pruning algorithm with a novel reward function, enabling controlled and efficient filter removal in CNNs.

Findings

Significant filter reduction achieved with minimal performance loss

Method validated on CNNs for recognition and segmentation tasks

Provides flexible tradeoff control between accuracy and model size

Abstract

Many state-of-the-art computer vision algorithms use large scale convolutional neural networks (CNNs) as basic building blocks. These CNNs are known for their huge number of parameters, high redundancy in weights, and tremendous computing resource consumptions. This paper presents a learning algorithm to simplify and speed up these CNNs. Specifically, we introduce a "try-and-learn" algorithm to train pruning agents that remove unnecessary CNN filters in a data-driven way. With the help of a novel reward function, our agents removes a significant number of filters in CNNs while maintaining performance at a desired level. Moreover, this method provides an easy control of the tradeoff between network performance and its scale. Per- formance of our algorithm is validated with comprehensive pruning experiments on several popular CNNs for visual recognition and semantic segmentation tasks.