Multi-objective Bayesian Optimization for Neural Architecture Search

0568613 - ÚI 2024 RIV CH eng C - Konferenční příspěvek (zahraniční konf.)
Vidnerová, Petra - Kalina, Jan
Multi-objective Bayesian Optimization for Neural Architecture Search.
Artificial Intelligence and Soft Computing. 21st International Conference, ICAISC 2022. Proceedings, Part I. Cham: Springer, 2023 - (Rutkowski, L.; Scherer, R.; Korytkowski, M.; Pedrycz, W.; Tadeusiewicz, R.; Zurada, J.), s. 144-153. Lecture Notes in Computer Science, 13588. ISBN 978-3-031-23491-0. ISSN 0302-9743.
[ICAISC 2022: International Conference on Artificial Intelligence and Soft Computing /21./. Zakopane (PL), 18.06.2022-22.06.2022]
Grant CEP: GA ČR(CZ) GA22-02067S
Institucionální podpora: RVO:67985807
Klíčová slova: Bayesian optimization * Multi-objective optimization * Neural architecture search * Number of parameters
Obor OECD: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
https://dx.doi.org/10.1007/978-3-031-23492-7_13

A novel multi-objective algorithm denoted as MO-BayONet is proposed for the Neural Architecture Search (NAS) in this paper. The method based on Bayesian optimization encodes the candidate architectures directly as lists of layers and constructs an extra feature vector for the corresponding surrogate model. The general method allows to accompany the search for the optimal network by additional criteria besides the network performance. The NAS method is applied to combine classification accuracy with network size on two benchmark datasets here. The results indicate that MO-BayONet is able to outperform an available genetic algorithm based approach.
Trvalý link: https://hdl.handle.net/11104/0339882