Coreset-Based Neural Network Compression

TL;DR

This paper introduces a coreset-based CNN compression method that significantly reduces model size and inference time without retraining, maintaining high accuracy and enabling domain adaptation.

Contribution

The authors present a novel coreset approach for CNN compression that requires no retraining and achieves state-of-the-art results across various architectures.

Findings

Achieves 832x reduction in memory footprint compared to AlexNet.

Maintains AlexNet-like accuracy after compression.

Reduces inference time significantly.

Abstract

We propose a novel Convolutional Neural Network (CNN) compression algorithm based on coreset representations of filters. We exploit the redundancies extant in the space of CNN weights and neuronal activations (across samples) in order to obtain compression. Our method requires no retraining, is easy to implement, and obtains state-of-the-art compression performance across a wide variety of CNN architectures. Coupled with quantization and Huffman coding, we create networks that provide AlexNet-like accuracy, with a memory footprint that is $832\times$ smaller than the original AlexNet, while also introducing significant reductions in inference time as well. Additionally these compressed networks when fine-tuned, successfully generalize to other domains as well.