Production Systems, the Utility problem in machine learning

Production systems (condition-action rule systems) are used extensively in investigating and building a wide range of AI systems. The issue of efficiency of these systems has long been of major concern. Much progress has in fact been made, beginning with the invention of the Rete network technology. However, the concerns about production system performance remain critical, with new challenges from areas such as learning systems, real-time systems, active database systems and so on.

Selected publications

• Tambe, M., and Rosenbloom, P. S. Investigating Production System Representations for Non-combinatorial Match. Artificial Intelligence 68(1), 1994.
• Acharya, A., and Tambe, M. Collection-oriented Match. In Proceedings of the Second International Conference on Information and Knowledge Management. November, 1993.
• Doorenbos, R., Tambe, M., and Newell, A. Learning 10,000 chunks: what's it like out there. In Proceedings of the National Conference on Artificial Intelligence. August, 1992.
• Acharya, A., Tambe, M., and Gupta, A. Implementation of Production Systems on Message Passing Computers: Simulation Results and Analysis. IEEE Transactions on Parallel and Distributed Computing 3(4):477-487, 1992.
• Harvey, W., Kalp, D., Tambe, M., McKeown, D., and Newell, A. The Effectiveness of Task-Level Parallelism for Production Systems. Journal of Parallel and Distributed Computing 13(4), December, 1991.
• Tambe, M., Newell A., and Rosenbloom, P. The Problem of Expensive Chunks and its Solution by Restricting Expressiveness. Machine Learning 5(3):299-348, 1990.
• Gupta, A., Tambe, M., Kalp, D., Forgy, C. L., and Newell, A. Parallel implementation of OPS5 on the Encore Multiprocessor: Results and analysis. International Journal of Parallel Programming 17(2):95-124, 1988.
• Tambe, M., Kalp, D., Gupta, A., Forgy, C.L., Milnes, B.G., and Newell, A. Soar/PSM-E: Investigating match parallelism in a learning production system. In Proceedings of the ACM/SIGPLAN Symposium on Parallel Programming: Experience with Applications, Languages, and Systems. July, 1988. (Also available in Rosenbloom, P., Laird, J. E., and Newell, A. ed., The Soar papers, MIT Press, Cambridge, MA).

  A Description of Rete

  See section 2.1 of this tech report

  Software

• CParaOPS5 a C-based OPS5 compiler can be ftp'ed from ftp.isi.edu. Change to directory /soar/CParaOPS/ Login as anonymous with user email address as the password.

  ---

  Milind Tambe
  tambe@isi.edu