Publication
Our paper, titled Beta Distribution-based XCS Classifier System, was published in the proceedings of IEEE Congress on Evolutionary Computation (CEC) 2022.
This work introduces a novel approach to Learning Classifier Systems by incorporating beta distribution for improved rule representation and adaptation.
Research Innovation
Traditional XCS (eXtended Classifier System) uses fixed condition representations that may not optimally capture the underlying data patterns. Our research introduces beta distribution-based representations that can dynamically adapt to the data characteristics.
Key Contributions
- Beta Distribution Integration: Novel use of beta distribution for condition representation in XCS
- Adaptive Rule Learning: Enhanced ability to learn and adapt rules for complex patterns
- Improved Performance: Demonstrated superior performance on benchmark problems
- Theoretical Foundation: Solid mathematical framework for the proposed approach
Technical Advancement
The beta distribution provides a flexible and mathematically principled way to represent fuzzy conditions in learning classifier systems, enabling better generalization and more effective learning in continuous domains.
Impact and Applications
- Pattern recognition and classification
- Adaptive control systems
- Data mining and knowledge discovery
- Evolutionary machine learning
Collaboration
This work represents a productive collaboration with my colleagues, demonstrating the power of combining different expertise areas in machine learning and evolutionary computation.
Authors: Hiroki Shiraishi, Yohei Hayamizu, Hiroyuki Sato, and Keiki Takadama (*Equal contribution)
Venue: IEEE Congress on Evolutionary Computation (CEC) 2022