Multi-Subspace Neural Network for Image Recognition

TL;DR

This paper introduces a multi-subspace neural network (MSNN) that combines CNN components with subspace concepts to improve image classification by better handling intra-class variability and providing diverse data interpretations.

Contribution

The novel integration of subspace concepts with CNN architecture in MSNN offers a new approach for robust feature extraction in image recognition tasks.

Findings

MSNN achieves competitive accuracy on MNIST and COIL-20 datasets.

The multi-subspace strategy enhances robustness to intra-class variability.

Experimental results outperform some existing methods.

Abstract

In image classification task, feature extraction is always a big issue. Intra-class variability increases the difficulty in designing the extractors. Furthermore, hand-crafted feature extractor cannot simply adapt new situation. Recently, deep learning has drawn lots of attention on automatically learning features from data. In this study, we proposed multi-subspace neural network (MSNN) which integrates key components of the convolutional neural network (CNN), receptive field, with subspace concept. Associating subspace with the deep network is a novel designing, providing various viewpoints of data. Basis vectors, trained by adaptive subspace self-organization map (ASSOM) span the subspace, serve as a transfer function to access axial components and define the receptive field to extract basic patterns of data without distorting the topology in the visual task. Moreover, the multiple-subspace strategy is implemented as parallel blocks to adapt real-world data and contribute various interpretations of data hoping to be more robust dealing with intra-class variability issues. To this end, handwritten digit and object image datasets (i.e., MNIST and COIL-20) for classification are employed to validate the proposed MSNN architecture. Experimental results show MSNN is competitive to other state-of-the-art approaches.