Convolutional Neural Network Compression through Generalized Kronecker Product Decomposition

TL;DR

This paper introduces a novel tensor decomposition method called GKPD to compress CNN layers, significantly reducing memory and computation while maintaining accuracy, and outperforming existing methods on standard benchmarks.

Contribution

We propose the Generalized Kronecker Product Decomposition (GKPD), a versatile plug-and-play module for CNN compression that surpasses current state-of-the-art tensor and other compression techniques.

Findings

GKPD outperforms Tensor-Train and Tensor-Ring decompositions.

GKPD achieves better compression with minimal accuracy loss.

Experimental results on CIFAR-10 and ImageNet validate effectiveness.

Abstract

Modern Convolutional Neural Network (CNN) architectures, despite their superiority in solving various problems, are generally too large to be deployed on resource constrained edge devices. In this paper, we reduce memory usage and floating-point operations required by convolutional layers in CNNs. We compress these layers by generalizing the Kronecker Product Decomposition to apply to multidimensional tensors, leading to the Generalized Kronecker Product Decomposition (GKPD). Our approach yields a plug-and-play module that can be used as a drop-in replacement for any convolutional layer. Experimental results for image classification on CIFAR-10 and ImageNet datasets using ResNet, MobileNetv2 and SeNet architectures substantiate the effectiveness of our proposed approach. We find that GKPD outperforms state-of-the-art decomposition methods including Tensor-Train and Tensor-Ring as well as other relevant compression methods such as pruning and knowledge distillation.