This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
Unsupervised Automata Learning via Discrete Optimization
Created by
Haebom
Author
Simon Lutz, Daniil Kaminskyi, Florian Wittbold, Simon Dierl, Falk Howar, Barbara K onig, Emmanuel Muller, Daniel Neider
Outline
In this paper, we propose a new framework for learning deterministic finite automata (DFA) from unlabeled data to overcome the limitations of automata learning methods that rely on conventional supervised learning environments. We prove the computational complexity of the problem of learning DFA from multiple sets of unlabeled words, and develop three learning algorithms based on constraint optimization. In addition, we propose a new regularization technique to improve the interpretability of DFA, and demonstrate its practicality in the field of unsupervised anomaly detection through a prototype implementation.
Takeaways, Limitations
•
Takeaways:
◦
Presenting new possibilities for automatic learning using unlabeled data
◦
Development of an efficient learning algorithm based on constraint optimization
◦
Proposing a new regularization technique to improve the interpretability of DFA
◦
Verification of practical applicability in unsupervised anomaly detection
•
Limitations:
◦
Limited scalability due to the computational complexity of the proposed problem.
◦
Experimental verification through prototype implementation requires more extensive experiments and verification on diverse datasets.
◦
Limiting performance evaluation for certain types of unlabeled data
