The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #94114

Title:The Baldwin Effect
Date: 1994
Abstract:It is widely accepted that learning cannot be transferred directly from an individual into its genetic material. However learning can play an important role in evolution by changing the shape of the space on which selection operates - this is called the "Baldwin effect". This project considers different approaches to modelling the Baldwin effect with Genetic Algorithms, applying them to a variety of problems. The effect is demonstrated successfully and confirmed to be particularly significant when the search space is not nicely structured. The limitations of the model are discussed and it is seen that the arguments for the significance of the Baldwin effect can be strengthened by using more realistic models of learning.

[Search These Pages] [DAI Home Page] [Comment]