SparseNet: A Sparse DenseNet for Image Classification

TL;DR

SparseNet introduces a sparsified DenseNet architecture that reduces connections, enhances efficiency, and incorporates an attention module, leading to improved or comparable performance with significantly fewer parameters and faster computation on image classification datasets.

Contribution

The paper proposes a novel sparsification method for DenseNet, reducing connection complexity from O(L^2) to O(L), and integrates an attention module to boost performance.

Findings

SparseNet outperforms state-of-the-art on CIFAR10 and SVHN.

SparseNet is 2.6 times smaller and 3.7 times faster than DenseNet.

Achieves comparable accuracy to DenseNet with fewer parameters.

Abstract

Deep neural networks have made remarkable progresses on various computer vision tasks. Recent works have shown that depth, width and shortcut connections of networks are all vital to their performances. In this paper, we introduce a method to sparsify DenseNet which can reduce connections of a L-layer DenseNet from O(L^2) to O(L), and thus we can simultaneously increase depth, width and connections of neural networks in a more parameter-efficient and computation-efficient way. Moreover, an attention module is introduced to further boost our network's performance. We denote our network as SparseNet. We evaluate SparseNet on datasets of CIFAR(including CIFAR10 and CIFAR100) and SVHN. Experiments show that SparseNet can obtain improvements over the state-of-the-art on CIFAR10 and SVHN. Furthermore, while achieving comparable performances as DenseNet on these datasets, SparseNet is x2.6 smaller and x3.7 faster than the original DenseNet.