v1.0.2

The functions were modified so they won't break if using neural networks trained with a skip layer. However, plotting and variable importance values may be unreliable. See the response in the issues section.