Grow-Push-Prune: aligning deep discriminants for effective structural network compression

TL;DR

This paper introduces Grow-Push-Prune, an iterative method for creating task-specific compact deep neural networks by aligning discriminants and pruning unnecessary neurons, demonstrated to outperform existing models on multiple datasets.

Contribution

It proposes a novel Grow-Push-Prune approach that combines discriminant alignment and neuron pruning, along with a network growing strategy, for effective model compression.

Findings

Achieves higher accuracy than comparable compact models on ImageNet.

Effectively reduces model complexity while maintaining or improving performance.

Demonstrates superiority over residual and other compact networks at similar sizes.

Abstract

Most of today's popular deep architectures are hand-engineered to be generalists. However, this design procedure usually leads to massive redundant, useless, or even harmful features for specific tasks. Unnecessarily high complexities render deep nets impractical for many real-world applications, especially those without powerful GPU support. In this paper, we attempt to derive task-dependent compact models from a deep discriminant analysis perspective. We propose an iterative and proactive approach for classification tasks which alternates between (1) a pushing step, with an objective to simultaneously maximize class separation, penalize co-variances, and push deep discriminants into alignment with a compact set of neurons, and (2) a pruning step, which discards less useful or even interfering neurons. Deconvolution is adopted to reverse 'unimportant' filters' effects and recover useful contributing sources. A simple network growing strategy based on the basic Inception module is proposed for challenging tasks requiring larger capacity than what the base net can offer. Experiments on the MNIST, CIFAR10, and ImageNet datasets demonstrate our approach's efficacy. On ImageNet, by pushing and pruning our grown Inception-88 model, we achieve more accurate models than Inception nets generated during growing, residual nets, and popular compact nets at similar sizes. We also show that our grown Inception nets (without hard-coded dimension alignment) clearly outperform residual nets of similar complexities.