LeTE

   

LeTE: Learnable Transformation-based Generalized Time Encoding

Code for the paper "Rethinking Time Encoding via Learnable Transformation Functions"

🔥Our paper has been accepted by ICML 2025.

You can also find the paper at arXiv.

Demo Experiments

You can run the demo experiments in the EventBasedMNIST.ipynb.

This notebook is corresponding to the experiments in the paper Section 4.1 Time as the Only Input.

Citation

@misc{chen2025lete,
      title={Rethinking Time Encoding via Learnable Transformation Functions}, 
      author={Xi Chen and Yateng Tang and Jiarong Xu and Jiawei Zhang and Siwei Zhang and Sijia Peng and Xuehao Zheng and Yun Xiong},
      year={2025},
      eprint={2505.00887},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2505.00887}, 
}

Poster