Examining the robustness of Physics-Informed Neural Networks to noise for Inverse Problems