Robust Training for Time-Series

Python Implementation of RObustTraining forTime-Series (RO-TS) for the paper: "Training Robust Deep Models for Time-Series Domain: Novel Algorithms and Theoretical Analysis" by Taha Belkhouja, Yan Yan, and Janardhan Rao Doppa.

Setup

pip install -r requirement.txt

By default, data is stored in experim_path_{dataset_name}. Directory can be changed in RO_TS.py

Obtain datasets

• The dataset can be obtained as .zip file from "The UCR Time Series Classification Repository".
• Download the .zip file and extract it it in UCRDatasets/{dataset_name} directory.
• Run the following command for pre-processing a given dataset while specifying if it is multivariate, for example, on SyntheticControl dataset

python preprocess_dataset.py --dataset_name=SyntheticControl --multivariate=False

The results will be stored in Dataset directory.

Run

• Example training run

python RO_TS.py --dataset_name=SyntheticControl --K=10 --rots_beta=5e-1 --rots_lambda=1e-2 --batch=11

• Example testing run

python test_RO_TS_model.py --dataset_name=SyntheticControl --rots_beta=5e-1 --rots_lambda=1e-2