Data Types as a More Ergonomic Frontend for Grammar-Guided Genetic Programming

Explore a 19-minute conference talk from ACM SIGPLAN on using data types as a more ergonomic frontend for Grammar-Guided Genetic Programming (GGGP). Delve into the proposed approach of embedding grammar as an internal Domain-Specific Language in the host language framework, offering the same expressive power as BNF and EBNF while leveraging existing tooling. Learn about Meta-Handlers, user-defined overrides of the tree-generation system, and how they extend object-oriented encoding with greater practicability and expressive power. Examine a Python implementation example and compare this approach to textual BNF-representations in terms of expressive power, ergonomics, and performance across 5 benchmarks against PonyGE2. Gain insights into the advantages of this method for improving Genetic Programming applications in Machine Learning, Optimization, and Software Engineering.

Intro
GENETIC PROGRAMMING
DEFINING THE SEARCH SPACE LASIGE
CHALLENGES IN GRAMMAR-GUIDED GP LASIGE
MOTIVATION
DATATYPE ENCODING
REAL-WORLD EXAMPLE
METAHANDLERS
EXPRESSIVE POWER
PERFORMANCE
ERGONOMICS
BONUS CONCLUSIONS