Anthony

Anthony Bucci

Ph.D., Computer Science, Brandeis University, 2007
B.S., Mathematics, Case Western Reserve University, 1995
Summary | Academic Activities | Publications | Tutorials | Resources | Contact Information
     

Summary
Coevolutionary algorithms typically explore domains in which no single evaluation function is present or known. For the purpose of selecting which individuals to maintain and vary, they instead rely on the outcomes of interactions between evolving entities. Exactly how outcomes are measured and how measurements are integrated into a selection decision has a strong impact on an algorithm's dynamics, specifically on whether capable individuals are found or interesting behavior arises. My research has explored the thesis that within interactive domains such as a set of game-playing strategies, there is an implicit, provably present and algorithmically-extractable set of yardsticks which might be called informative dimensions. Somewhat surprisingly, these dimensions are not just theoretical curiosities; experiments suggest they have a meaning in terms of human-interpretable features of a domain. Furthermore, a characterizing property of interactive domains is that they have more than one dimension. Various dynamic effects like cycling or overspecialization can then be explained in terms of how an algorithm trades off between dimensions. Experiments suggest that algorithms which have been sensitized to the presence of informative dimensions show better dynamic behavior.
     

Academic Activities
Here are some things I have been or will be doing:
Talks/Presentations
  • Greythumb Boston (Jun 2009): Invited Talk
    Title: Why Might Coevolution Be Good at Worst-case Optimization?
  • Tufts University Department of Computer Science (Apr 2009): Class Lecture, cs150
    Title: Coevolutionary Algorithms
  • GE Global Research (Dec 2008): Invited Talk
    Title: When Performing Is Not Enough: Informativeness Incentives in Coevolutionary Algorithms
  • Greythumb Boston (Aug 2008): Invited Talk
    Title: TBD (video!)
  • Genetic and Evolutionary Computation Conference (Jul 2008): Tutorial
    Title: Advanced Tutorial on Coevolution (presented with Sevan Ficici)
  • Dagstuhl Theory of Evolutionary Algorithms Seminar (Feb 2008): Invited Talk
    Title: Solution Concepts and the Subdomain Representation
  • Tufts University Department of Computer Science (Dec 2007): Invited Talk
    Title: When Performing Is Not Enough: Informativeness Incentives in Coevolutionary Algorithms
  • Dissertation Defense Talk (1PM 20 July 2007)
    Title: Emergent Geometric Organization and Informative Dimensions in Coevolutionary Algorithms
Randomonia
I was on the program committee of FOGA 2011, FOGA 2009, FOGA 2007 and the GECCO 2007 coevolution track. I have been on the program committees of the GECCO 2006, GECCO 2005, GECCO 2004 and GECCO 2003 coevolution tracks.
I am reviewing or have reviewed papers for the following journals: IEEE Transactions on Systems, Man, and Cybernetics, IEEE Transactions on Neural Networks, IEEE Transactions on Evolutionary Computation, Artificial Life and Machine Learning Journal
I have been known to tutor mathematics, even category theory.
I have worked on computational biology with Dr. Jason Johnson, then at the Pfizer Discovery Technology Center in Cambridge, MA. We presented the work as a poster at ISMB 2001 in July 2001. I have also applied insights from genetic programming and coevolutionary algorithms to the computational chemistry problem of solubility prediction. This work was a strategic alliance with Pfizer, under the sponsorship of Dr. Jason Hughes.
     

Dissertation
Bucci, A. (2007). Emergent Geometric Organization and Informative Dimensions in Coevolutionary Algorithms. Presented to the faculty of the Michtom School of Computer Science at Brandeis University, 20 Jul 2007.
     

Book Chapters
Popovici, E., Bucci, A., Wiegand, P. and De Jong E.D. (2010). Coevolutionary Principles. Chapter to appear in Handbook of Natural Computing, Springer-Verlag.
De Jong, E.D. and Bucci, A. (2008). Objective Set Compression: Test-Based Problems and Multi-Objective Optimization. Chapter in Multi-Objective Problem Solving from Nature: From Concepts to Applications, Springer-Verlag.
     

Publications and Posters
Popovici, E., Winston, E., and Bucci, A. (2011). On the Practicality of Optimal Output Mechanisms for Co-optimization Algorithms. Accepted at FOGA 2011.
Stuermer, P., Bucci, A., Branke, J., Funes, P. and Popovici, E. (2009). Analysis of Coevolution for Worst-case Optimization. Accepted at GECCO 2009, Genetic Algorithms Track.
[nominated for best paper award]
Bucci, A. and Pollack, J.B. (2007). Thoughts on Solution Concepts. Accepted at GECCO 2007, Coevolution Track.
De Jong, E.D. and Bucci, A. (2006). DECA: Dimension Extracting Coevolutionary Algorithm. Accepted at GECCO 2006, Coevolution Track.
Bucci, A. and Pollack, J.B. (2005). On Identifying Global Optima in Cooperative Coevolution. Accepted at GECCO 2005, Coevolution Track.
Bucci, A., Pollack, J.B. and De Jong, E.D. (2004). Automated Extraction of Problem Structure. Accepted at GECCO 2004, Coevolution Track.
[nominated for best paper award]
Bucci, A. and Pollack, J.B. (2003). Focusing versus Intransitivity: Geometrical Aspects of Co-evolution. GECCO 2003, Coevolution Track.
Bucci, A. and Pollack, J.B. (2003). A Mathematical Framework for the Study of Coevolution. FOGA 7: Proceedings of the Foundations of Genetic Algorithms Workshop. (link is a draft version; see Morgan Kaufmann for book details).
Errata PS PDF
Bucci, A. and Pollack, J.B. (2002). Order-theoretic Analysis of Coevolution Problems: Coevolutionary Statics. GECCO 2002, Workshop on Understanding Coevolution: Theory and Analysis of Coevolutionary Algorithms.
Bucci, A. and Johnson, J.M. (2001). Neural network and genetic algorithm identification of coupling specificity and functional residues in G protein-coupled receptors. ISMB 2001 Poster Session: Alignment Techniques.
     

Tutorials
Ficici, S.G. and Bucci, A. (2007). Advanced Tutorial on Coevolution. GECCO 2007.
Ficici, S.G. and Bucci, A. (2006). Introductory Tutorial on Coevolution. GECCO 2006.
     

Resources
The DEMO Lab at Brandeis University, directed by Professor Jordan Pollack.
The EC Lab at George Mason University, directed by Professor Ken De Jong.
Computational Intelligence and Coevolution Group
A Google group on coevolution, part of the IEEE Evolutionary Computation Technical Committee.
Larry Bull
Paul Darwen
Edwin de Jong
Edwin de Jong has a nice coevolution page.
Sevan Ficici
Sevan Ficici works on understanding coevolution using Evolutionary Game Theory (EGT). He has forwarded the ideas of Pareto Coevolution and solution concept, both of which have been important in clarifying what coevolutionary algorithms do. See his dissertation for details.
Ludo Pagie
Mitch Potter's Dissertation
Mitch Potter introduced cooperative coevolution while at Ken De Jong's EC Lab. CCEAs have been used to solve difficult problems and have become an active research area.
Kenneth Stanley
Richard Watson
Richard Watson's SEAM algorithm has begun making inroads into the question of adapting representations in evolutionary algorithms. His numbers game work has provided a suite of test problems for coevolution which illuminate what can go wrong at the evaluation stage of a coevolutionary algorithm.
R. Paul Wiegand
Paul Wiegand works on EGT models of cooperative coevolution.
IlliGAL Blog
A blog maintained by the folks at the Illinois Genetic Algorithms Lab.
EMOO Repository
Resources on Evolutionary Multi-Objective Optimization (EMOO). Maintained by Professor Carlos Coello Coello.
Google Scholar is going after the niche that CiteSeer has dominated for several years.
     

Contact Information
Anthony Bucci
Computer Science Department MS018
Brandeis University
Waltham MA 02454
(781) 736 3366 office/lab
(781) 736 2700 department office
(781) 736 2741 department fax
 abucci ( ^ ) cs o brandeis o edu
     
Charles Darwin We can allow satellites, planets, suns, universe, nay whole systems of universes, to be governed by laws, but the smallest insect, we wish to be created at once by special act.