The evolution of PAMELA

Abstract

This paper describes the results of a$ve-year research program on production systems carried out in Vienna by a large engineering group at the Alcatel-Elin research laboratory. In the course of our investigations, a real-time expert system shell, PAMELA, was developed. We describe features of PAMELA and discuss ways in which the matching algorithm can be speeded up. It was found that the single processor version is one of the most efficient forward chainers available worldwide. Since single processor enhancements have been widely explored, we concentrated our efforts on multiple processor architectures to provide further per$ormance gain. EfJicient exploitation of parallelism in production systems requires suitable algorithms for load-balancing, without simultaneously increasing organisational or communication overheads. With this in mind, we developed a transputer-based architecture which serves as a testbedfor our new parallel matching algorithm. The execution speeds of various expert systems based on PAMELA are presented.

Efficiency Mechanisms for a Class of Blackboard Systems

“…Furthermore, by storing the matching information in the network, pattern-matching networks eliminate redundant computation from matching the same patterns that occur in different productions. Every production system uses some variation of RETE [26] as an efficient activation mechanism.…”

Collection oriented match

“…These systems have traditionally used more powerful match algorithms like Rete and its derivatives [4,11] and Treat [22].…”

Investigating production system representations for non-combinatorial match

“…In a production with multiple conditions, a cascading effect of such branching leads to an exponential match effort, i.e., a k-search tree with a number of tokens O(#WMEscondiCwnf) (751. This cascading effect is also referred to as the cross-product effect [30, 311, since it yields the cross-product of the bindings for the variables 6 .…”

“…However. there are still cases in which the direct computation of the dominar" instantiation is sufficiently complex that it is preferable to compute all of the instantiations first and then select the dominant one rather than trying to compute it directly [6,7].…”

“…Outside of Soar, Chalasani and Altmann [19891 have pointed out that the knowledge representation scheme adopted by Theo, a frame-based architecture for problem solving and learning [56], already corresponds to the unique-attribute representation. Unique-attributes have also been mapped over (in an extended form) to the PAMELA expert system [6,9], to the K production system [141, and simple unique-attribute set structures have been mapped over to Prolog control rules [181.…”