I am interested in understanding and creating systems that learn through experience to obtain rewards. This seems to me the essence of intelligence: "Intelligent behavior is to be repeatedly successful in satisfying your needs in diverse, observably different, situations on the basis of past experience"*.
Human beings and other animals are such systems, as are the systems studied in the branch of artificial intelligence called reinforcement learning. My own research has concentrated on learning classifier systems (LCS), evolutionary reinforcement learners conceived by J. H. Holland in the 1970's that through continuing development, are beginning to display aspects of intelligent behavior.
A basic introduction to learning classifier systems (2 pages, PDF) is here. A comprehensive introduction, review, and roadmap to the field (as of 2008) is here. A history of LCS to 2014 is here. A chapter on XCS and XCSF from the Springer Handbook of Computational Intelligence (2015) is here. The LCS Wikipedia page is here. A 12-minute YouTube video, "Learning Classifier Systems in a Nutshell", is here.
Areas of progress and current LCS research include: accurate generalization, learning to obtain deferred rewards, learning in non-Markov ("aliased") environments, learning under noisy inputs, learning continuous-valued actions, learning relational concepts, and learning inverse dynamics. Applications include data inference, compound ("hyper") heuristics, on-line function approximation, financial prediction, autonomous agent ("animat") robotics, salient object detection, and detection and visualization of gene interactions.
"The animat path to AI" is an introduction to this view of intelligence. "Classifier fitness based on accuracy" has the basics of my recent work on classifier systems. My latest** work is "Coevolution of pattern generators and recognizers". The "Colleagues' web pages" link above is a doorway to recent LCS publications and their authors.
Prediction Dynamics® is my research and consulting organization***. I am available for consulting in areas that relate to the research, particularly prediction from data and software and robotic agents. A resume (older) is here. A 2011 ACM SIGEVOlution interview is here. I can be contacted at firstname.lastname@example.org
* P. J. van Heerden, The Foundation of Empirical Knowledge, Wistik, The Netherlands, 1968, p. 10.
** Recently I conceived and successfully applied the idea of using a genetic algorithm to evolve
Strategies in the new financial field of Temporal Portfolio Theory (TPT).
*** Prediction Dynamics® is a registered trademark of Stewart W. Wilson.