Optimizing Convolutional Neural Networks for Embedded Systems by Means of Neuroevolution

TL;DR

This paper introduces a neuroevolution approach to automatically optimize CNN architectures for embedded systems, focusing on reducing power consumption and complexity while maintaining classification accuracy.

Contribution

The proposed method evolves CNNs considering both error rate and complexity, enabling efficient deployment on resource-constrained embedded devices.

Findings

Achieved competitive accuracy on MNIST and CIFAR-10 benchmarks.

Reduced power consumption by utilizing fixed point operations during inference.

Demonstrated the effectiveness of neuroevolution in optimizing CNNs for embedded systems.

Abstract

Automated design methods for convolutional neural networks (CNNs) have recently been developed in order to increase the design productivity. We propose a neuroevolution method capable of evolving and optimizing CNNs with respect to the classification error and CNN complexity (expressed as the number of tunable CNN parameters), in which the inference phase can partly be executed using fixed point operations to further reduce power consumption. Experimental results are obtained with TinyDNN framework and presented using two common image classification benchmark problems -- MNIST and CIFAR-10.