Interpreting Deep Learning Models for Knowledge Tracing