Carleton University
Comparing Parameter Estimation and State Prediction Performance of Physics Informed Neural Networks in Relation to Bayesian Inference
Abstract
dc:description.abstractThis thesis investigates two physics informed neural network (PINN) approaches for two mechanistic problems via the performance metrics of model parameter estimation and system state prediction. The PINNs are compared to Bayesian inference (BI) tested under similar conditions. The first PINN studied is the loss-based PINN, which embeds physics into a standard feedforward neural network within the loss function of this network. The second PINN investigated is the hybrid-recurrent physics informed neural network, which uses a recurrent neural network (RNN) that embeds physics in RNN architecture. The BI used as a point of comparison uses a Transitional Markov Chain Monte Carlo algorithm to estimate the posterior distributions. The applications studied are a compartmental model for an infectious disease, and a mass-spring-damper model. It is found that BI performs better in ideal data conditions and the PINNs perform better in weak data conditions.
Degree
thesis:*- Name thesis:degree_name
- Master of Applied Science (M.App.Sc.)
- Level thesis:degree_level
- Master's
- Discipline thesis:degree_discipline
- Engineering, Civil
- Grantor dc:publisher
- Carleton University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Pantano, Michael Francesco
Subjects
dc:subject × 3Rights
dc:rights- Statement dc:rights
-
- Copyright © 2025 the author(s). Theses may be used for non-commercial research, educational, or related academic purposes only. Such uses include personal study, distribution to students, research and scholarship. Theses may only be shared by linking to the Carleton University Institutional Repository and no part may be copied without proper attribution to the author; no part may be used for commercial purposes directly or indirectly via a for-profit platform; no adaptation or derivative works are permitted without consent from the copyright owner.
- Language dc:language.iso
- en