Time-Transformer

Pytnon implementation paper "Time-Transformer: Integrating Local and Global Features for Better Time Series Generation" (SDM24).

Jupyter Notebook "tutorial" provide a tutorial for training and evaluating with different metrics (using "sine_cpx" dataset). FID score are calculated with "fid_score" in ts2vec, directly using model "TS2Vec".

The model is built with "tensorflow2", please check the "requirement.txt" and decide which package you need to run the model.

If you find this model useful and put it in your publication, we encourage you to add the following references:

@misc{liu2023timetransformer,
      title={Time-Transformer: Integrating Local and Global Features for Better Time Series Generation}, 
      author={Yuansan Liu and Sudanthi Wijewickrema and Ang Li and Christofer Bester and Stephen O'Leary and James Bailey},
      year={2023},
      eprint={2312.11714},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}