Evaluating Gam-Like Neural Network Architectures For Interpretable Machine Learning