Abstract
Information Theory, Fitness Landscapes and Evolutionary Adaptation
After a brief introduction as to why someone like me works for the phone company, this talk will summarise ongoing research into evolutionary adaptation and the structure of fitness landscapes using information theory.
Starting with a quick overview of ideas central to artificial evolution, I will motivate this very theoretical research in terms of the need to turn artificial evolution into a better box of tools for practical engineering applications. After critically evaluating alternative information theoretic upper bounds, due to Worden and to Chater, on the rate at which a population may adapt to its environment, I relate the speed of evolution to the structure and information theoretic properties of the fitness landscape. I introduce additional information-based tools, such as algorithmic information content and logical depth, and conclude with a discussion of how they might be applied to further improve our understanding of evolutionary adaptation.
The talk will be broadly accessible, with each technical tool carefully introduced and explained.
|
