Welcome to the Supplementary Materials Repository for our work Explaining Time Series Classifiers trhough Meaningful Perturbation and Optimisation. This repository contains additional resources that support our research findings.
To run the full experiment. Download the UEA datasets and put it into the folder /dataset/Data/MultivariateUEA/... you need to process the original datasets and save it in a numpy file.
Train_Classifier.py -- Train the classier to be explained. Train_GenModel.py -- Train the generative model for realistic inputs generation. main_train_E2Gan.py and main_rain_BRITS.py train two time series imputation models.
Explaining.py --- explain Trained models. Evaluation.py -- evaluation the results.