Black-box models for TERP interpretation

TERP is a post-hoc interpretation scheme for explaining black-box AI predictions. TERP works by constructing a linear, local interpretable model that approximates the black-box in the vicinity of the instance being explained. TERP determines the accuracy-interpretability trade-off by introducing and using the concept of interpretation entropy.This data repository contains the three trained machine learning models: VAMPnets, Vision Transformers models (ViT) - pre-trained (model.ckpt)+ fine-tuned (best-model.ckpt) + fine-tuned_data_randomized (bad-model.ckpt), attention-based bi-directional LSTM) trained on molecular dynamics simulation trajectory of alanine dipeptide, facial attributes of celebrities (CelebA), and Antonio Gulli’s (AG’s) news corpus respectively. The simulated trajectory (dihedral angles) for the molecular dynamics simulation is also provided.