Capturing Actionable Dynamics with Structured Latent Ordinary Differential Equations (UAI 2022)

This repository contains the Pytorch code to replicate experiments in our paper Capturing Actionable Dynamics with Structured Latent Ordinary Differential Equations accepted at Conference on Uncertainty in Artificial Intelligence (UAI 2022):

@inproceedings{chapfuwa2022capturing,
  title={Capturing Actionable Dynamics with Structured Latent Ordinary Differential Equations},
  author={Chapfuwa, Paidamoyo and Rose, Sherri and Carin, Lawrence and Meeds, Edward and Henao, Ricardo},
  booktitle={The 38th Conference on Uncertainty in Artificial Intelligence},
  year={2022}
}

Model

Prerequisites

The code is implemented with the following dependencies:

• Python 3.8.10
• Torch 2.3.0
• Pyro 1.9.0
• Additional python packages can be installed by running:

pip install -r requirements.txt

Data

We consider the following datasets:

• Human Viral Challenge
• Cardiovascular System (CVS)
• Synthetic Biology

Model Training

• To train the data specific SLODE models run:

  • training_challenge.py
  • training_cvs.py
  • training_proc.py

• The data specific hyper-parameters settings can be found at:

  • config_challenge.py
  • config_cvs.py
  • config_proc.py

Metrics and Visualizations

• Once the networks are trained and the results are saved, we visualize the data specific key results:
  • challenge_eval_folds.ipynb for cross validation experiments and challenge_eval_folds_subject_final.ipynb for subject specific qualitative results
  • cvs_eval_final.ipynb
  • sbio_eval_folds_final.ipynb for cross validation experiments and sbio_eval_heldout_final.ipynb for zero-shot heldout device