Efficient Training of Physics-Informed Neural Networks through Joint Optimization of Collocation and Experimental Data Points

This work introduces PINNACLE, the first algorithm that jointly optimizes the selection of all training point types for Physics-Informed Neural Networks (PINNs), including collocation points for enforcing PDEs and initial/boundary conditions, as well as experimental data points. PINNACLE automatically adjusts the proportion of collocation point types as training progresses to boost PINN performance.