PINN Training using Biobjective Optimization: The Trade-off between Data Loss and Residual Loss