Abstract:
Learnable Evolution Model (LEM) is an evolutionary computation methodology that applies hypothesis formulation and instantiation to create new individuals. Initial study has shown that LEM significantly outperforms standard evolutionary computation methods in terms of evolution length on selected benchmark optimization problems. This paper presents initial results from handling constrained optimization problems in LEM. Constraints are classified as instantiable, which can be handled directly during instantiation process, and general, which cannot be directly instantiated. The latter can be handled by applying three different methods presented in this paper.
