GM-Net: Learning Features with More Efficiency

TL;DR

This paper introduces GM-Net, a deep CNN architecture with a novel merging strategy and basic units that improve parameter efficiency and classification performance across multiple image datasets.

Contribution

The paper proposes a new GM-Net architecture with Basic Units and a merging strategy that enhances feature reuse and reduces parameters, outperforming existing methods.

Findings

GM-Net achieves superior accuracy on MNIST, CIFAR-10, CIFAR-100, and SVHN.

The proposed architecture significantly reduces network parameters.

Extensive experiments validate the effectiveness of GM-Net.

Abstract

Deep Convolutional Neural Networks (CNNs) are capable of learning unprecedentedly effective features from images. Some researchers have struggled to enhance the parameters' efficiency using grouped convolution. However, the relation between the optimal number of convolutional groups and the recognition performance remains an open problem. In this paper, we propose a series of Basic Units (BUs) and a two-level merging strategy to construct deep CNNs, referred to as a joint Grouped Merging Net (GM-Net), which can produce joint grouped and reused deep features while maintaining the feature discriminability for classification tasks. Our GM-Net architectures with the proposed BU_A (dense connection) and BU_B (straight mapping) lead to significant reduction in the number of network parameters and obtain performance improvement in image classification tasks. Extensive experiments are conducted to validate the superior performance of the GM-Net than the state-of-the-arts on the benchmark datasets, e.g., MNIST, CIFAR-10, CIFAR-100 and SVHN.