Abstract:

A typical evolutionary system acts in a set-and-forget fashion: parameters and settings are specified, then the system is run independently of the user. However in many cases, the results are unacceptable to the user: parameter bounds are unacceptable to the user, genetic programming generates solutions of unacceptable form, etc. So the system in fact operates in a loop: when the user deems solutions unacceptable, this additional knowledge is added to the system, usually as a constraint, and the system is re-run. This process is repeated until an acceptable solution is obtained.

This process is often successful, acting to clarify the user's knowledge about acceptable solutions. But it incorporates a subtle problem: only the constraints that are in fact violated are recorded as explicit user biases. The user may well have additional biases that would have been specified if necessary, but these are not available for inspection.

This talk will explore the issue of user biases in evolutionary computation, with particular attention to previous experience in genetic algorithms and genetic programming, and planned extensions to interactive evolutionary computation.

Bio Sketch:

Bob McKay was originally trained as a pure mathematician in the Australian National University and the University of Bristol. He is currently a professor at the Dept of Computer Science and Engineering, College of Engineering, Seoul National University. He's a member of the Structural Complexity Laboratory, researching Genetic Programming and related fields. Dr. McKay obtained his BSc Hons in Pure Mathematics from Australian National University and his PhD in Theory of Computation from Bristol. He's been the keynote/plenary/invited speakers in many academic conferences as well as playing major roles in various international conferences. Dr. McKay has been researching the field of evolutionary computation since the 1990s, at first in the University of New South Wales and subsequently in Seoul National University. He has participated in IES symposia since the first in Australia in 1997. He is an Associate Editor of IEEE Trans Evolutionary Computation, Editorial Board member of Genetic Programming and Evolvable Machines (Springer), and Associate Editor of KES Journal (IOS Press).