CANTS-GP: A Nature-Inspired Metaheuristic for Graph Based Genetic Programs
Chapter
Overview
Overview
Abstract
Graph-based genetic programs built from arithmetic primitives (e.g., addition,multiplication, trigonometric functions) offer inherently interpretable models forcomplex tasks. This work introduces CANTS-GP, an extension on Continuous Ant-Based Neural Topology Search (CANTS), which is a nature-inspired metaheuristicthat constructs computational graphs with trainable arithmetic functions rather thanopaque neural units. CANTS-GP employs a multi-colony framework: simulated antagents traverse an unbounded continuous search space guided by pheromone trails,and colonies periodically exchange evolving parameters—pheromone evaporationrates and the number of foraging ants—via a particle swarm optimization inspiredinformation sharing. Ant-generated paths are consolidated into flexible, acycliccomputational graphs through the DBSCAN clustering algorithm and depth firstsearch based cycle removal. Key innovations include adaptive colony evaporationcontrol, dynamic movement heuristics, and multi-colony evolutionary strategies.We evaluate CANTS-GP on six benchmark time-series forecasting tasks, demon-strating consistent performance gains over state-of-the-art methods while yieldingtransparent models whose arithmetic structures provide model interpretability.
