Pseudo-Physics-Informed Neural Operators: Enhancing Operator Learning from Limited Data