Revisit Lmser and its further development based on convolutional layers

TL;DR

This paper revisits the Lmser network, extending it with convolutional layers for image tasks, and demonstrates its effectiveness in recognition, reconstruction, and recall through experiments.

Contribution

It introduces a deep convolutional Lmser network, enhancing the original auto-encoder based Lmser for modern image processing applications.

Findings

Lmser functions are validated in image tasks

Convolutional Lmser shows promising performance

Effective in recognition, reconstruction, and association recall

Abstract

Proposed in 1991, Least Mean Square Error Reconstruction for self-organizing network, shortly Lmser, was a further development of the traditional auto-encoder (AE) by folding the architecture with respect to the central coding layer and thus leading to the features of symmetric weights and neurons, as well as jointly supervised and unsupervised learning. However, its advantages were only demonstrated in a one-hidden-layer implementation due to the lack of computing resources and big data at that time. In this paper, we revisit Lmser from the perspective of deep learning, develop Lmser network based on multiple convolutional layers, which is more suitable for image-related tasks, and confirm several Lmser functions with preliminary demonstrations on image recognition, reconstruction, association recall, and so on. Experiments demonstrate that Lmser indeed works as indicated in the original paper, and it has promising performance in various applications.