Researchers have developed an Evolution Strategy (ES) to construct partial spread bent functions, which are Boolean functions with maximum nonlinearity. The ES was tested with two encoding methods, one using generator matrices and the other using spread sets, successfully finding these functions for up to 10 variables. An ablation study indicated that mutation is crucial, while population size is not, and a greedy heuristic outperformed the ES on challenging instances. Further analysis explored quality-diversity variants of the ES to find functions with low algebraic degrees, revealing that the evolved bent functions exhibited higher 2-ranks and degrees compared to those generated by a baseline Genetic Programming optimizer. AI
IMPACT This research introduces novel methods for evolving complex mathematical functions, potentially impacting future AI research in areas requiring advanced representation techniques.
RANK_REASON Academic paper detailing a new method for constructing specific types of mathematical functions. [lever_c_demoted from research: ic=1 ai=0.7]
Read on arXiv cs.NE (Neural & Evolutionary) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →