Proceedings of the
9th International Symposium for Geotechnical Safety and Risk (ISGSR)
25 – 28 August 2025, Oslo, Norway
Editors: Zhongqiang Liu, Jian Dai and Kate Robinson
Proposal For AIC in a Reduced-Order Model Based on Proper Orthogonal Decomposition
1Engineering Seismology Group, Earthquake Disaster Prevention Engineering Department, Port and Airport Research Institute (PARI), National Institute of Marine, Port and Aviation Technology (MPAT), Japan.
2Structural Design Division, ECOH Corporation, Japan.
3AIS Lab. (Advanced Infrastructure Systems), Department of Civil Environmental Engineering, Tohoku University, Japan
4Big Data Technology Group, Infrastructure Digital Transformation Engineering Department, Port and Airport Research Institute (PARI), National Institute of Marine, Port and Aviation Technology (MPAT),Japan
5National Institute of Land and Infrastructure Management (NILIM), Ministry of Land, Infrastructure, Transport and Tourism (MLIT), Japan
ABSTRACT
In this study, we propose an equation for the Akaike Information Criterion (AIC) as an index to quantitatively evaluate the generalization performance of a reduced-order model (ROM) based on proper orthogonal decomposition (POD). The ROM proposed by Otake et al. (2021) has primarily been validated using known input data, and the evaluation of its generalization performance for unknown input data has been an issue. In this study, we incorporate a linear Gaussian state-space model into the ROM by Otake et al. (2021) and present a method to quantitatively evaluate its generalization performance using AIC through parameter estimation with the Expectation-Maximization algorithm.
Keywords: Akaike Information Criterion (AIC), Proper Orthogonal Decomposition (POD), Reduced-OrderModel (ROM), Lenear Gaussian State-Space Model (LGSSM).
