gep

Gene Expression Programming (GEP) is a type of evolutionary computation that evolves computer programs or mathematical expressions. It uses a population of fixed-length, linear chromosomes that encode expression trees via a genotype-phenotype mapping. Each chromosome is composed of one or more genes; each gene encodes a sub-expression using a function set and a terminal set. The phenotype is assembled by linking these sub-expressions into a complete form such as a parse tree or an arithmetic expression, ensuring syntactically correct outputs.

During evolution, standard genetic operators such as mutation and recombination are applied to the linear chromosomes. GEP also employs specialized transposition operators that move or copy gene fragments to produce new genomes. Selection is fitness-based, guiding the population toward expressions that better fit the data or objective function.

Compared with traditional genetic programming, GEP separates genotype from phenotype more clearly and guarantees syntactic validity of offspring, which simplifies the search process and often speeds convergence. It can reduce code bloat and facilitate more scalable exploration of solution spaces.

Common applications include symbolic regression, data-driven modeling, system identification, time-series forecasting, and rule-based classification. Researchers also adapt GEP with extensions such as multi-gene representations, strongly typed GEP, and domain-specific function sets to improve performance on particular tasks.

Because the acronym GEP can refer to other terms in different domains, some texts may use gep in lowercase to indicate unrelated concepts; in computing and evolutionary computation, however, GEP almost always denotes Gene Expression Programming.