MONCAE: Multi-Objective Neuroevolution of Convolutional Autoencoders

TL;DR

This paper introduces MONCAE, a neuroevolutionary approach using a hypervolume indicator to optimize convolutional autoencoder architectures, achieving significant image compression while maintaining classification accuracy.

Contribution

It presents the first use of a hypervolume indicator in neural architecture search for autoencoders, advancing AutoML techniques for image compression.

Findings

Images compressed by over 10x

Maintained classification performance

Accelerated AutoML pipeline for image tasks

Abstract

In this paper, we present a novel neuroevolutionary method to identify the architecture and hyperparameters of convolutional autoencoders. Remarkably, we used a hypervolume indicator in the context of neural architecture search for autoencoders, for the first time to our current knowledge. Results show that images were compressed by a factor of more than 10, while still retaining enough information to achieve image classification for the majority of the tasks. Thus, this new approach can be used to speed up the AutoML pipeline for image compression.