May 2002
I have just completed my PhD thesis -
And I am now a Postdoctoral Fellow at the
2102 Biological Laboratories
Harvard University
16 Divinity Avenue
Cambridge, MA 02138, USA
tel: (617) 495-1568
fax: (617) 496-5854
rwatson@oeb.harvard.edu
Previously.....
PhD student in the Dynamical and Evolutionary Machine OrganizationMy advisor is Jordan Pollack
MSc., Knowledge Based Systems,
University of Sussex, U.K., 1996.
B.A., Computing with Artificial Intelligence,
University of Sussex, U.K., 1990.Research interests
Research history
Publications
Contact Information
Research Interests
thesis proposal.My research is centered on the development of an algorithmic understanding of symbiosis in nature.
En route, I have passed through some different topic areas like Collective Evolutionary Robotics, a lot of Genetic Algorithms research, and some simple models of 'Symbiotic Scaffolding'. But, to date, my research culminates in the Symbiogenic Evolutionary Adaptation Model (SEAM).
Collective Evolutionary Robotics: In some early thesis work Sevan Ficici and I developed a new methodology for evolving controllers using a population of robots that breed with one another 'on-the-fly' in the task environment. We call this "Embodied Evolution".Genetic Algorithms (GAs): A large part of my effort on GAs has been directed at the development of a GA test problem that I call "Hierarchical-if-and-only-if" (H-IFF). It has features that make it ideally suited for the GA with recombination and (maximally) hard for any kind of mutation-based algorithm. With H-IFF I have investigated the operation of the GA especially with respect to the Building-Block Hypothesis and recombination operators. (see FOGA 2001 paper)
Symbiotic Scaffolding: In 1999/2000 I have produced a model of how an organism can acquire the characteristics of a symbiont without genetic transfer. (See papers for ECAL99 and ALife7.)
Symbiogeneic Evolutionary Adaptation Model (SEAM): My motivating research question is: What is symbiosis good for? (in the algorithmic sense). In particular, I am driven to understand the natural phenomenon of "symbiogenesis": the genesis of new species via genetic integration of symbionts. This composition of pre-adapted parts is a fundamentally different source of innovation from the Darwinian accumulation of random variations, and it has been instrumental in several of the major transitions in evolutionary history (for example, the evolution of eukaryotes which compose all plants and animals). In my view, this process is algorithmically powerful because it enables the discovery and composition of 'building-blocks' in much the same sense as that meant in GA theory. But, through my GA work, I have come to understand the important differences between symbiotic composition and sexual recombination in evolutionary algorithms. Most recently, I have developed an intergrated model of various concepts that have been explored in previous work. It is the first of my papers that brings together work on symbiosis with Genetic Algorithms. (see PPSN 2000, ECAL 2001). See also symbiosis for a discussion of how the previous papers fit together with this one.
The following items each say something about what Im interested in a general sense:
Thesis proposal.
Symbiosis view of my thesis work
Research directions spring 1999 - big picture
Embodied Evolution
Research directions 1999
GECCO'99 PhD student workshop (one page summary of GA work upto May 1999)In my spare time, I like to discuss language acquisition with someone who is nominally a UG believer Laura Dominguez
Research History
PhD Research in progresslog of thoughts on research in progress 2001
log of thoughts on research in progress 2000
log of thoughts on research in progress 1999
log of thoughts on research in progress since arriving at Brandeis upto end 1998Masters Research, University of Sussex, U.K., 1995-96.
Symbiogenesis: the role of symbiogenesis in natural evolution (its pervasiveness). The application of this priciple to practical engineering problems.Independent Research, Townsville, North Queensland, Australia, 1994-95.
Architecture for neural networks enabling tractable learning from reinforcement, layers of successive indirection.Industrial Research, British Telecommunications, London, 1990-92.
Virtual documents. Multiple organisations of information fragments provide several virtual documents presenting the information in a hierarchy best suited to the point of view of the user.Undergraduate Research, University of Sussex, U.K., 1987-90.
A system that learns to do basic arithmetic: Constructing rules from ambiguous examples, identifying correct interpretations via usage and marking. Utilises rule-strength dynamics that model fact-recal based on psycological models of recency and frequency.
Publications
Please see the 'official list' For a list of publications by subject area see:
for abstracts and full ps/ pdf of published papers
Please see the 'official list'
Thesis pageFor details on ICCS 2002 presentation please see:
Watson RA & Pollack JB, 2001, "A computational Model of Symbiotic Composition in Evolutionary Transitions", accepted for Biosystems special issue on Evolvability. (pdf version) <= this version is not final.Publications not in official list (yet)
Knowles, J.D. and Watson, R.A. (2002) On the Utility of Redundant Encodings in Mutation-based Evolutionary Search. To appear in J.J. Merelo et al. (Eds.) Proceedings of the 7th International Conference on Parallel Problem Solving from Nature (PPSN-VII). Copyright Springer-Verlag. Download: Postscript, PDF
Knowles, J.D., Watson, R.A., Corne, D.W. (2001) Reducing Local Optima in Single-Objective Problems by Multi-objectivization. In Proceedings of the First International Conference on Evolutionary Multi-criterion Optimization (EMO'01), pp. 269--283, copyright Springer-Verlag.
Other Stuff
Watson RA & Pollack JB, 2000, "Combination and Recombination in Genetic Algorithms", technical report CS-00-209, Dept. Computer Science, Brandeis University.(pdf version)
Watson RA, Hornby GS, & Pollack JB, 1998, "Modeling Building-Block Interdependency", PPSN-V, 1998.
(N.B. the version of this paper appearing in the PPSN proceedings has an error in the legend of figure 2. Specifically, uniform and 2-point cross-over are erroneously swapped. This on-line version is correct :)
C code for the HIFF test function is provided here.Watson RA, 1996 "A study of the Effect of Group Formation on Evolutionary Search", MSc dissertation, University of Sussex, 1996. (but the charts are missing - let me know if you are interested.)
"Evolution and Problem Decomposition", (summary of ongoing dissertation research for GECCO-99 PhD student workshop), Workshop Program, 1999, pp.416-417. (pdf version)
"Hierarchically-Consistent GA Test Problems: Summary and Additional Results", GECCO-99 Late Breaking Papers, 1999, pp 292-297. (pdf version)
"Embodied Evolution ", SAB98 poster (letter page format), 1998.
Brandeis contact information
See New web page for Harvard Contact infoRichard A Watson
Mail stop 018,
Volen Center for Complex Systems
Computer Science Department
Brandeis University
415 South Street,
Waltham, MA 02254-9110, USA(781) 736-3366 office phone
(781) 736-2700 dept secretary
(781) 736-2741 fax
richardw@cs.brandeis.edu
Go back up.
Mail comments to webmaster@cs.brandeis.edu.count started 30/08/00
Richard A. Watson Richard A Watson R. A. Watson R A Watson R. Watson R Watson Watson, R. A. Watson R.A. Watson R A