A simple version of Pytorch implementation of Multi-task Adversarial Spatial-Temporal Networks for Crowd Flow Prediction published in CIKM2020 where Self-attention module was replaced by a simple attention mechanism.
Numpy
scipy
Python 3.6+
Pytorch 1.0+
Matplotlib(Optional)
Please cite the following paper if you use this repository in your reseach.
@inproceedings{wang2020multi,
title={Multi-task Adversarial Spatial-Temporal Networks for Crowd Flow Prediction},
author={Wang, Senzhang and Miao, Hao and Chen, Hao and Huang, Zhiqiu},
booktitle={Proceedings of the 29th ACM International Conference on Information \& Knowledge Management},
pages={1555--1564},
year={2020}
}
The code might be a little messy. For any question, feel free to contact:
Hao Miao: haomiao96@gmail.com
Senzhang Wang: szwang@nuaa.edu.cn