Continual-Learning Physics-Informed Neural Networks for Parameterized Partial Differential Equations