Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Acceleration
This work addresses the challenge of accelerating deep CNNs for image classification tasks, offering a novel approach that advances state-of-the-art in model compression.
The paper tackles the problem of inefficient filter pruning in convolutional neural networks by proposing a new method that prunes redundant filters via geometric median, achieving over 52% FLOP reduction on ResNet-110 with a 2.69% accuracy improvement on CIFAR-10 and over 42% FLOP reduction on ResNet-101 without accuracy drop on ILSVRC-2012.
Previous works utilized ''smaller-norm-less-important'' criterion to prune filters with smaller norm values in a convolutional neural network. In this paper, we analyze this norm-based criterion and point out that its effectiveness depends on two requirements that are not always met: (1) the norm deviation of the filters should be large; (2) the minimum norm of the filters should be small. To solve this problem, we propose a novel filter pruning method, namely Filter Pruning via Geometric Median (FPGM), to compress the model regardless of those two requirements. Unlike previous methods, FPGM compresses CNN models by pruning filters with redundancy, rather than those with ''relatively less'' importance. When applied to two image classification benchmarks, our method validates its usefulness and strengths. Notably, on CIFAR-10, FPGM reduces more than 52% FLOPs on ResNet-110 with even 2.69% relative accuracy improvement. Moreover, on ILSVRC-2012, FPGM reduces more than 42% FLOPs on ResNet-101 without top-5 accuracy drop, which has advanced the state-of-the-art. Code is publicly available on GitHub: https://github.com/he-y/filter-pruning-geometric-median