Improving the Explain-Any-Concept by Introducing Nonlinearity to the Trainable Surrogate Model | IEEE Conference Publication | IEEE Xplore