- author = "Ondrej Galeta",
- title = "Transformer-Guided Mutation in Cartesian Genetic Programming for Approximate Multiplier Design",
- booktitle = "Student Conference on Innovation, Technology and Science in IT",
- year = "2026",
- editor = "Petr Veigend",
- address = "Brno, Czech Republic",
- month = "5 " # may,
- keywords = "genetic algorithms, genetic programming, Cartesian Genetic Programming",
-
URL = "
https://excel.fit.vutbr.cz/submissions/2026/039/39.pdf",
- size = "3 pages",
- abstract = "Cartesian Genetic Programming (CGP) for approximate circuit design suffers from poor scalability due to expensive evaluations. This work proposes a transformer-guided mutation operator to accelerate the design of approximate multipliers. A BERT-based model predicts where and how to mutate circuit representations, supported by dataset filtering, augmentation, and a fallback to standard mutation. Results on EvoApprox8b show faster convergence and improved solutions over standard CGP for some error thresholds. The approach improves CGP speed of convergence, creates new potentially patentable designs, and demonstrates the potential of combining evolutionary design with machine learning.",
-
notes = "See also
\cite{galeta2026geneticprogrammingtransformerbasedmutation}
Studentska Konference Inovaci, Technologii a Vedy v IT Excel@FIT http://excel.fit.vutbr.cz/
Faculty of Information Technology, Brno University of Technology",
