Farmer, J. Doyne, and Norman H. Packard. “Evolution, games, and learning: Models for adaptation in machines and nature. An introduction to the proceedings of the CNLS Conference, Los Alamos, May 1985.” Physica D: Nonlinear Phenomena 22.1-3 (1986): vii-xii.
URL
At first glance, “models for adaptation in machines and nature” might seem an unusual topic for a physics journal. What does it have to do with nonlinear phenomena? As illustrated by the papers in this volume, however, serious models of adaptation are inevitably nonlinear. Adaptive behavior is an emergent property, which spontaneously arises through the interaction of simple components. Whether these components are neurons, amino acids, ants, or bit strings, adaptation can only occur if the collective behavior of the whole is qualitatively different from that of the sum of the individual parts. This is precisely the definition of nonlinear.
The purpose of this volume (and the conference that inspired it) is to bring together the study of adaptive processes in nature and their implementation in artificial systems, exploring what these different approaches have in common and what they have to learn from one another. Adaptation is the common thread linking the topics in the title, and in this proceedings. Although nonlinear dynamics is a prerequisite for adaptation, work on this subject has mainly been done by specialists in different fields, often without full awareness of recent developments in nonlinear dynamics. Our goal in publishing this work in Physica D is to introduce the fertile problems surrounding adaptation to the nonlinear dynamics community, with the hope that it can provide a focal point for work done in diverse disciplines. This approach has been quite successful in the study of chaos, which in a certain sense is the related opposite of adaptation and self-organization. Adaptation is one of the most complex and least understood types of dynamical behavior, and it is high time that it be addressed directly by nonlinear dynamicists.
The three topics stated in the title of the conference were inspired by a common interest in them, combined with the feeling that there are deeper connections between them than meet the eye. Evolution, after all, is one of the best exampies of a process capable of spontaneous learning and design. Somehow nature begins with very simple building blocks and engineering takes place automatically as complex organisms develop. Elucidation of the underlying principles of adaptation in biological evolution may well lead to better methods of achieving machine learning and artificial intelligence. Conversely, adaptive schemes for machine learning may provide a simple context in which to investigate the mathematical basis of biological evolution.