Preprocessing in Attractor Neural Networks

TL;DR

This paper investigates the role of preprocessing, specifically Fourier transform, in attractor neural networks for invariant pattern recognition, highlighting the importance of phase information over amplitude.

Contribution

It demonstrates that the effectiveness of Fourier-based preprocessing depends heavily on the amount of stored information, emphasizing the significance of phase in recognition.

Findings

Fourier transform preprocessing's performance varies with stored information amount.

Phase information is more critical than amplitude in recognition accuracy.

Preprocessing impact is significant in neural network pattern recognition.

Abstract

Preprocessing the input patterns seems the simplest approach to invariant pattern recognition by neural networks. The Fourier transform has been proposed as an appropriate and elegant preprocessor. Nevertheless, we show in this work that the performance of this kind of preprocessor is strongly affected by the number of stored informations. This is because the phase of the Fourier transform plays a more important role than the amplitude in the recognition process.