Surrogate Modeling for Efficient Evolutionary Multi-Objective Neural Architecture Search in Super Resolution Image Restoration

Sergio Sarmiento-Rosales, Jesús Llano García, Jesús Falcón-Cardona, Raúl Monroy, Manuel Casillas del Llano, Víctor Sosa-Hernández

2024

Abstract

Fully training each candidate architecture generated during the Neural Architecture Search (NAS) process is computationally expensive. To overcome this issue, surrogate models approximate the performance of a Deep Neural Network (DNN), considerably reducing the computational cost and, thus, democratizing the utilization of NAS techniques. This paper proposes an XGBoost-based surrogate model to predict the Peak-Signal-to-Noise Ratio (PSNR) of DNNs for Super-Resolution Image Restoration (SRIR) tasks. In addition to maximizing PSNR, we also focus on minimizing the number of learnable parameters and the total number of floating-point operations. We use the Non-dominated Sorting Genetic Algorithm III (NSGA-III) to tackle this three-objective optimization NAS problem. Our experimental results indicate that NSGA-III using our XGBoost-based surrogate model is significantly faster than using full or partial training of the candidate architectures. Moreover, some selected architectures are comparable in quality to those found using partial training. Consequently, our XGBoost-based surrogate model offers a promising approach to accelerate the automatic design of architectures for SRIR, particularly in resource-constrained environments, decreasing computing time.

Paper Citation

in Harvard Style

Sarmiento-Rosales S., Llano García J., Falcón-Cardona J., Monroy R., Casillas del Llano M. and Sosa-Hernández V. (2024). Surrogate Modeling for Efficient Evolutionary Multi-Objective Neural Architecture Search in Super Resolution Image Restoration. In Proceedings of the 16th International Joint Conference on Computational Intelligence - Volume 1: ECTA; ISBN 978-989-758-721-4, SciTePress, pages 242-249. DOI: 10.5220/0012949000003837

in Bibtex Style

@conference{ecta24,
author={Sergio Sarmiento-Rosales and Jesús Llano García and Jesús Falcón-Cardona and Raúl Monroy and Manuel Casillas del Llano and Víctor Sosa-Hernández},
title={Surrogate Modeling for Efficient Evolutionary Multi-Objective Neural Architecture Search in Super Resolution Image Restoration},
booktitle={Proceedings of the 16th International Joint Conference on Computational Intelligence - Volume 1: ECTA},
year={2024},
pages={242-249},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012949000003837},
isbn={978-989-758-721-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 16th International Joint Conference on Computational Intelligence - Volume 1: ECTA
TI - Surrogate Modeling for Efficient Evolutionary Multi-Objective Neural Architecture Search in Super Resolution Image Restoration
SN - 978-989-758-721-4
AU - Sarmiento-Rosales S.
AU - Llano García J.
AU - Falcón-Cardona J.
AU - Monroy R.
AU - Casillas del Llano M.
AU - Sosa-Hernández V.
PY - 2024
SP - 242
EP - 249
DO - 10.5220/0012949000003837
PB - SciTePress