Chris Rosin

I am currently with Parity Computing. You can email me there at crosin@cparity.com. This page is somewhat out of date.

I completed my Ph.D. in the CSE Department at UCSD in June, 1997. My advisor was Rik Belew. I was then a LJIS postdoc at The Scripps Research Institute, working in Art Olson's Lab. Since January, 1999, I have worked as a research scientist with Parity Computing Inc..

Interests

Artificial intelligence, machine learning, intelligent text processing, computational biology, genetic algorithms, coevolution, computational learning theory, game learning, neural networks, complex systems, cellular automata.

Thesis and Postdoctoral Research

For many problems, performance of candidate solutions needs to be measured on a suite of test cases. The choice of test cases can be an important factor in the quality of solutions found. Adversarial problems are those for which it is possible to search for good test cases as well as good solutions. Examples include games (strategies must be tested against strong opponents), controller design (test against difficult plants), and grammar induction (test on strings that reveal unusual features of the language). I am exploring theoretical aspects of adversarial problems, and coevolutionary heuristics for solving them. My postdoctoral research focus was the application of this work to the problem of drug resistance in drug design. Coevolution is used to search for drug designs effective across a broad range of possible resistance mutations. This method relies on computational models of drug interaction; I am also working on theoretical approaches to choosing experiments to perform, for the purpose of gathering additional data to improve these models during search.

Curriculum Vitae

You can view my C.V. as plain text.

Publications

• Rosin, C.D. (2000). "Sample complexity of model-based search." Journal of Computer and System Sciences 60:278-301.
• Rosin, C.D., Belew, R.K., Walker, W.L., Morris, G.M., Olson, A.J., Goodsell, D.S. (1999). Coevolution and subsite decomposition for the design of resistance-evading HIV-1 protease inhibitors. Journal of Molecular Biology 287(1):77-92.
• Rosin, C.D., Belew, R.K., Morris, G.M., Olson, A.J., and Goodsell, D.S. (1999). "Coevolutionary analysis of resistance-evading HIV-1 protease inhibitors." Proceedings of the National Academy of Sciences 96(4):1369-1374.
• Rosin, C.D. (1998). "Sample complexity of model-based search." Proceedings of the Eleventh Annual Conference on Computational Learning Theory. ACM.
• Rosin, C.D., Belew, R.K., Morris, G.M., Olson, A.J., and Goodsell, D.S. (1998). "Computational coevolution of antiviral drug resistance." Artificial Life 4:41-59.
• Rosin, C.D., Belew, R.K., Morris, G.M., Olson, A.J., and Goodsell, D.S. (1998). "Computational coevolution of antiviral drug resistance." Proceedings of the Sixth International Conference on Artificial Life. MIT Press.
• "Coevolutionary Search Among Adversaries" Christopher D. Rosin. Ph.D. Dissertation, University of California, San Diego, 1997.
• "A Comparison of Global and Local Search Methods in Drug Docking", Christopher D. Rosin, R. Scott Halliday, William E. Hart, and Richard K. Belew. Submitted version. Final version in the Proceedings of the Seventh International Conference on Genetic Algorithms. An older version of this paper is available as Technical Report #CS97-522.
• Technical Report #CS96-491: "New methods for Competitive Coevolution", Christopher D. Rosin and Richard K. Belew. Final version in Evolutionary Computation 5:1.
• ICGA 95 paper (submitted version): Two new methods are used to improve the performance of competitive co-evolution on several games, including Go. "Methods for Competitive Co-evolution: Finding Opponents Worth Beating", Christopher D. Rosin and Richard K. Belew. Final version in Proceedings of the Sixth International Conference on Genetic Algorithms. L.J. Eshelman, editor.
• COLT 96 paper (submitted version): A computational learning theoretic model of game learning is described, and sufficient conditions are given for polynomial-time learnability of perfect strategies. "A Competitive Approach to Game Learning", Christopher D. Rosin and Richard K. Belew. Final version in Proceedings of the Ninth Annual ACM Conference on Computational Learning Theory.
• An older unpublished extended abstract on Competitive Learning, a theoretical model of learning in a competitive environment.
• Niebur, E., Koch, C., and Rosin, C.D. (1993). "An oscillation-based model for the neuronal basis of attention." Vision Research 33:2789-2802.

Links

• Machine Learning in Games
• Live Alife Page
• Computer Go