close fullscreen

Hyperon AI Algorithms+MOSES (Meta-Optimizing Semantic Evolutionary Search)

help edit space_dashboard
Responsible: Nil Geisweiller

Meta-Optimizing Semantic Evolutionary Search — breeds compact, interpretable programs using evolutionary search with Elegant Normal Form to constrain the combinatorial search space.

  • GitHub
</> Example Implementation
Evolutionary program synthesis
Demonstrates MOSES evolving interpretable boolean programs using semantic-aware evolutionary search.
; MOSES evolves compact programs ; Elegant Normal Form constrains search !(moses-evolve (fitness-fn classification-accuracy) (training-data health-records) (max-generations 500) (representation boolean-formula)) ; => (AND (OR $x1 $x3) ; (NOT $x7)) ; Interpretable, compact solution

Technical Deep Dive: MOSES Full — Elegant Normal Form, deme-based search, MORK integration roadmap, and implementation findings.