TEN-DM

May 12, 2026 ยท View on GitHub

This repo contrains the code for our paper TEN-DM: Topology-Enhanced Diffusion Model for Spatio-Temporal Event Prediction.

Environment Setup

  • Tested OS: Linux
  • Python >= 3.7
  • PyTorch == 1.7.1
  • Tensorboard

Dependencies:

  1. Install PyTorch 1.7.1 with the correct CUDA version.
  2. Use the pip install -r requirements. txt command to install all of the Python modules and packages used in this project.

Model Training

Data should be one of JP_Earthquake|COVID19|Thefts|311Service|US_Earthquake|.
Use the following command to train TEN-DM:

python run.py --dataset $dataset

Example run command of training TEN-DM on JP_Earthquake dataset:

python run.py --dataset JP_Earthquake

You can optionally specify additional hyperparameters to control training:

python run.py --dataset $dataset --timesteps $timesteps --samplingsteps $samplingsteps --batch_size $batch_size --total_epochs $total_epochs --loss_type $loss_type

The trained models are saved in ModelSave/.
The logs are saved in logs/.
The test results are saved in ModelResult/.