This project develops a novel hybrid framework integrating evolutionary algorithms with advanced machine learning models to revolutionize digital circuit design. Current methodologies face significant challenges: evolutionary approaches suffer from inefficient blind operators and expensive evaluations, while ML-based approaches lack hardware-specific training data. Our framework addresses these limitations through ML-guided evolutionary operators, structure-aware surrogate models, and specialized techniques for emerging technologies and verification requirements. The synergistic combination leverages evolutionary algorithms' exploration capabilities with ML's pattern recognition power. We will validate our approach through diverse case studies including approximate accelerators, verification-optimized designs, medical signal classifiers, and superconducting circuits, advancing automated circuit design for both current and emerging technologies.

Keywords
Digital circuit;Design Automation;Evolutionary Algorithms;Machine-Learning

26-22525M

English

Mrázek Vojtěch, Ing., Ph.D. - principal person responsible
Hurta Martin, Ing. - fellow researcher
Zachariášová Marcela, Ing., Ph.D. - fellow researcher

Department of Computer Systems
- responsible department (27.3.2025 - not assigned)
Department of Computer Systems
- beneficiary (27.3.2025 - 31.12.2030)

Responsibility: Mrázek Vojtěch, Ing., Ph.D.