The code in this repository for paper "Difference-Guided Representation Learning Network for Multivariate Time-Series Classification" accepted by IEEE Transactions on Cybernetics.
- python 2.7
- tensorflow 1.14.0
The 18 MTS (Multivariate Time Series) benchmark data sets can download from link. They are collected from different repository, such as UCI, UCR and so on. These benchmark data sets come from various fields and have various input ranges and different numbers of classes, variables, and instances. The .p files are lists of three numpy arrays with [samples, lables, original_lengths]. Meanwhile,
samples.shape = (number of instances, time length, number of variables)
labels.shape = (number of instances,)
original_lengths.shape = (number of instances,)
You can create a folder dataset and put the datasets in the folder.
You can run the command
python AD.py
to test the model on the ASD dataset.