Search papers, labs, and topics across Lattice.
This paper introduces a continual-learning physics-informed neural network (CL-PINN) that addresses the challenges of training parameterized PINNs by treating different parameter instances as related tasks and learning them sequentially. The method incorporates Bayesian optimization for active parameter selection, dynamic loss weighting, and sparse replay to enhance knowledge retention and mitigate overfitting. Evaluations on multiple benchmarks demonstrate that CL-PINN achieves superior and more balanced accuracy compared to traditional fixed-sampling methods, significantly reducing computational queries while maintaining generalization across parameter domains.
CL-PINN outperforms traditional methods by balancing accuracy and efficiency, making it a game-changer for solving parameterized PDEs without observational data.
Physics-informed neural networks (PINNs) incorporate governing equations into neural-network training and can approximate PDE solutions without requiring large observational datasets. Parameterized PINNs (ParamPINNs) further take physical parameters as inputs, allowing a single model to represent a family of PDE solutions over a parameter domain. Existing ParamPINNs, however, still face inefficient training, uneven accuracy across parameters, and overfitting to a limited set of sampled parameter tasks, which can impair generalization to unsampled parameters. To address these issues, we propose a continual-learning physics-informed neural network (CL-PINN), which treats PDE instances at different parameter values as related tasks and learns them sequentially. CL-PINN combines Bayesian-optimization-based active parameter selection, task-wise dynamic loss weighting, sparse physics-constrained replay, and an optional parameter subnetwork to improve task allocation and knowledge retention under bounded active-task capacity. It requires no observational data and is designed to solve parameterized PDEs over relatively broad parameter domains under limited computational resources. Multi-seed evaluations on five benchmarks, including one continuous function and four parameterized PDEs, show that Bayesian selection substantially reduces objective-loss queries relative to grid-greedy search, while sparse replay mitigates forgetting of earlier tasks. Under the prescribed within-case resource protocols, CL-PINN generally provides higher and more balanced solution accuracy than fixed-sampling and grid-greedy baselines. CL-PINN offers a practical route toward learning PDE solutions that generalize across physical parameters and has the potential to support reusable physics-informed surrogates for large-scale engineering parameter studies.
