The Utilities of Evolutionary Multiobjective Optimization for Neural Architecture Search – An Empirical Perspective

Liu，Xukun

10.1007/978-981-99-1549-1_15

2023

1865-0929

1865-0937

Communications in Computer and Information Science

1801 CCIS

179-195

Evolutionary algorithms have been widely used in neural architecture search (NAS) in recent years due to their flexible frameworks and promising performance. However, we noticed a lack of attention to algorithm selection, and single-objective algorithms were preferred despite the multiobjective nature of NAS, among prior arts. To explore the reasons behind this preference, we tested mainstream evolutionary algorithms on several standard NAS benchmarks, comparing single and multi-objective algorithms. Additionally, we validated whether the latest evolutionary multi-objective optimization (EMO) algorithms lead to improvement in NAS problems compared to classical EMO algorithms. Our experimental results provide empirical answers to these questions and guidance for the future development of evolutionary NAS algorithms.

deep learning Evolutionary multiobjective optimization neural architecture search

第一 ; 通讯

英语

EI

20232414217616

Arts computing ; Deep learning ; Multiobjective optimization

Ergonomics and Human Factors Engineering:461.4 ; Data Processing and Image Processing:723.2 ; Optimization Techniques:921.5

2-s2.0-85161431285

Scopus

Southern University of Science and Technology,Shenzhen,China

Liu，Xukun. The Utilities of Evolutionary Multiobjective Optimization for Neural Architecture Search – An Empirical Perspective[C],2023:179-195.