The Interplay between Evolution and Learning in Dynamic Environments: Computational Metaphor of Adaptive Algorithms

This thesis aims to explore the potential of combining evolution and learning in adaptive systems. It focuses on the interplay between social and asocial learning in dynamic environments, examining the situations in which socially-learned, individually-learned, or genetically-specified behaviours are more advantageous. A computational approach using a genetic algorithm to model evolution and neural networks for learning is employed to study the dynamics of the interaction between these adaptive systems. The results show that allowing for both individual and social learning improves the performance of evolutionary algorithms in dynamic environments. The thesis presents fundamental research questions, employs computational models to address them, and proposes future directions for the development of artificial intelligence based on evolution, learning, and culture.