Convolution with Logarithmic Filter Groups for Efficient Shallow CNN

TL;DR

This paper introduces a logarithmic filter grouping method for shallow CNNs that improves accuracy and parameter efficiency in classification tasks, especially suitable for mobile applications.

Contribution

The paper proposes a novel logarithmic filter grouping scheme that captures filter distribution nonlinearity, enhancing shallow CNN performance and efficiency.

Findings

Outperforms uniform filter grouping in accuracy.

Reduces parameter size in shallow CNNs.

Effective in mobile application scenarios.

Abstract

In convolutional neural networks (CNNs), the filter grouping in convolution layers is known to be useful to reduce the network parameter size. In this paper, we propose a new logarithmic filter grouping which can capture the nonlinearity of filter distribution in CNNs. The proposed logarithmic filter grouping is installed in shallow CNNs applicable in a mobile application. Experiments were performed with the shallow CNNs for classification tasks. Our classification results on Multi-PIE dataset for facial expression recognition and CIFAR-10 dataset for object classification reveal that the compact CNN with the proposed logarithmic filter grouping scheme outperforms the same network with the uniform filter grouping in terms of accuracy and parameter efficiency. Our results indicate that the efficiency of shallow CNNs can be improved by the proposed logarithmic filter grouping.