Harmonic Networks: Integrating Spectral Information into CNNs

TL;DR

This paper introduces harmonic blocks that incorporate spectral information via the Discrete Cosine Transform into CNNs, improving classification performance on several datasets.

Contribution

It proposes a novel spectral filter integration method using harmonic blocks to enhance CNNs' feature extraction capabilities.

Findings

Harmonic CNNs achieve comparable or better accuracy than traditional CNNs.

Harmonic blocks improve spectral feature learning in CNNs.

Validated on small NORB, CIFAR10, and CIFAR100 datasets.

Abstract

Convolutional neural networks (CNNs) learn filters in order to capture local correlation patterns in feature space. In contrast, in this paper we propose harmonic blocks that produce features by learning optimal combinations of spectral filters defined by the Discrete Cosine Transform. The harmonic blocks are used to replace conventional convolutional layers to construct partial or fully harmonic CNNs. We extensively validate our approach and show that the introduction of harmonic blocks into state-of-the-art CNN baseline architectures results in comparable or better performance in classification tasks on small NORB, CIFAR10 and CIFAR100 datasets.