LUIaard

Implementation of 'Linearly Updating Intervals' (LUI) for Robust Anytime Learning of Markov Decision Processes (NeurIPS 2022).

---

Developers

This implementation was build in Java on top of PRISM (probabilistic model checker), see http://www.prismmodelchecker.org for more.

Main developers:

• Marnix Suilen
• Thiago D. Simão
• David Parker

Contributers:

• Nils Jansen

---

Usage

To run the experiments from the paper:

• enter the PRISM directory and type cd prism then make.
• run python run.py $(seq -s \ 0 99) --cpus 2 to run all experiments presented in the paper for seeds 0..99 in parallel on 2 CPUs.

The output is saved in the folder prism/results in seperate files/folders for each experiments and seed.

The prism/notebooks directory contains python scripts to plot the results.

Datasets (models)

The models can be found in the prism/models directory.

---