-
Machine Learning Based SAT Solver for Dual Degree Project treating SAT as a classification problem
- SATZilla - Source Code from the official SATZilla Website to extract features
- The features are used to populate and prepare a feature based DB
- This feature DB is then trained on examples from the SAT Competion
- Current Models obtain accuracy of ~80% on average using MLP ALgorithms -Work currently in progress...
-
Important References
- Devlin, David, and Barry O’Sullivan. "Satisfiability as a classification problem." Proc. of the 19th Irish Conf. on Artificial Intelligence and Cognitive Science. 2008.
- Xu, Lin, et al. "SATzilla: portfolio-based algorithm selection for SAT." Journal of Artificial Intelligence Research (2008): 565-606.