UNIVERSITY OF HERTFORDSHIRE COMPUTER SCIENCE RESEARCH COLLOQUIUM presents "Evolutionary Robustness" Prof. W. B. Langdon, (University College London) 11 March 2025 13:00 -14:00 Room C408 Everyone is Welcome to Attend Refreshments will be provided Abstract: Even in stable conditions life can retain its ability for continued evolutionary improvement for more than 80,000 generations. Instead of 38 years, Genetic Programming (GP) experiments can be run to a million generations in a few weeks. Information theory explains why genetic programming populations converge and why the rate of fitness improvement falls as huge evolved programs become more robust to crossover. Mutation testing of large C++ programs shows that real software can also be robust to many source code changes. As with genetic programming, there is a tendency for deeply nested code to be more robust. For sustained innovation, we suggest architectures composed of many small evolvable units which are intimate with their environment, so that the impact of crossovers and mutations has only a short way to travel before impacting fitness. Biography: Bill Langdon is an Honorary Research Professor in the CREST group in University College London's Computer Science department. He has extensive expertise in Genetic Programming (including three books on Genetic Programming), and more recently Genetic Improvement. He has done post docs and made research visits to Holland (CWI, La Poutre, Vitanyi), Germany (Wolfgang Banzhaf), Sweden (Peter Nordin) and Canada as well as the UK (Birmingham Uni, Essex Uni, King's College London, and University College London). Before returning to university, Bill worked for 13 years in industry (real time control, pipelines, office automation, networking). --------------------------------------------------- Hertfordshire Computer Science Research Colloquium http://cs-colloq.cs.herts.ac.uk