Robust Simulation-based Inference under missing data via Neural Processes (RISE)

Yogesh Verma | Ayush Bharti | Vikas Garg

The code repository for the paper Robust Simulation-based Inference under missing data via Neural Processes, ICLR 2025.

The code is built on the top of the following repo. Please follow the guidelines in the reference repository to install the requisite packages.

Ref: Learning Robust Statistics for Simulation-Based Inference Under Model Misspecification

Citation

If you find this helpful in your research, please consider citing the following paper:

@inproceedings{
verma2025robust,
title={Robust Simulation-Based Inference under Missing Data via Neural Processes},
author={Yogesh Verma and Ayush Bharti and Vikas Garg},
booktitle={The Thirteenth International Conference on Learning Representations},
year={2025},
url={https://openreview.net/forum?id=GsR3zRCRX5}
}

Note: Please change the device (GPU or CPU) in the files accordingly, depending on which you are running.

Baselines

• We utilized the official code implementation of Simformer to run the baseline.
• For NPE-NN (Lueckmann et. al 2017), we followed code implementation here.
• For Wang et. al 2024, we utilize train the NPE with the binary mask indicator (as described in official paper), and can be run by

python -u train_glm_wang/_glu_wang.py --degree degree --type mcar

Running RISE

To run the task on GLU and GLM dataset for mcar/mnar under a certain degree run,

python -u train_glm/glu.py --degree degree --type mcar/mnar

Note: We are also constantly updating and revising the repo to make it more adaptable in a general way, and finidng bugs and removing them and modifying certain parts.

License

This code is released under the MIT License.