# Bayesian Structural Learning for an Improved Diagnosis of Cyber-Physical Systems

arXiv: https://arxiv.org/abs/2104.00987 · HTML: https://mlsanity.com/paper/2104.00987/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 5 of its 70 references that are in the map.
- First posted on arXiv 2 Apr 2021.

## Builds on (most cited first, 5 of 5)

- [XGBoost: A Scalable Tree Boosting System](https://mlsanity.com/paper/1603.02754/) (2016)
- [Learning Bayesian Networks: The Combination of Knowledge and Statistical Data](https://mlsanity.com/paper/1302.6815/) (2013)
- [A hybrid algorithm for Bayesian network structure learning with application to multi-label learning](https://mlsanity.com/paper/1506.05692/) (2015)
- [Decision-Theoretic Troubleshooting: A Framework for Repair and Experiment](https://mlsanity.com/paper/1302.3563/) (2013)
- [Explainable Restricted Boltzmann Machines for Collaborative Filtering](https://mlsanity.com/paper/1606.07129/) (2016)

## Abstract (arXiv)

The diagnosis of cyber-physical systems aims to detect faulty behaviour, its root cause and a mitigation or even prevention policy. Therefore, diagnosis relies on a representation of the system's functional and faulty behaviour combined with observations of the system taken at runtime. The main challenges are the time-intensive building of a model, possible state-explosion while searching for the root cause and interpretability of the results. In this paper we propose a scalable algorithm tackling these challenges. We use a Bayesian network to learn a structured model automatically and optimise the model by a genetic algorithm. Our approach differs from existing work in two aspects: instead of selecting features prior to the analysis we learn a global representation using all available information which is then transformed to a smaller, label-specific one and we focus on interpretability to facilitate repairs. The evaluation shows that our approach is able to learn a model with equal performance to state-of-the-art algorithms while giving better interpretability and having a reduced size.