Learning to Prune Filters in Convolutional Neural Networks
This addresses the issue of high redundancy and resource consumption in CNNs for computer vision applications, offering a data-driven method to control the tradeoff between performance and network scale.
The paper tackles the problem of reducing the size and computational cost of convolutional neural networks (CNNs) by introducing a learning algorithm that prunes unnecessary filters, achieving significant filter removal while maintaining performance levels.
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.