This is a fork of the github repository for the NeurIPS 2021 paper "CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation".
Additional contribution is to wrap this algorithm for car following trajectory dataset.
Please install the packages in requirements.txt
Download car following trajectory dataset as directed by this paper.
python exe_trajectory.py --testmissingratio [missing ratio] --nsample [number of samples]'visualize_examples_trajectory.ipynb' is a notebook for visualizing results.
Position, speed and acceleration with no missing values:
with missing values, our trained model will impute the values with its mean and variance visualized: (performance yet under tuning)
The CSDI paper:
@inproceedings{tashiro2021csdi,
title={CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation},
author={Tashiro, Yusuke and Song, Jiaming and Song, Yang and Ermon, Stefano},
booktitle={Advances in Neural Information Processing Systems},
year={2021}
}