TOPOLOGICALLY TRAVERSING THE RETE NETWORK
Mark W. Perlin · Applied Artificial Intelligence · 1990
The appropriate selection of data and control abstractions is essential in understanding and implementing algorithms. The RETE match algorithm (Forgy, 1979), which underlies the-efficiency of modern forward chaining expert systems, has a reputation for extreme difficulty. In this article, we present different, more appropriate data and control abstractions for RETE. In particular, we demonstrate why topological traversal is the correct control mechanism for computing conjunctive match with RETE networks. It improves both the clarity and the ease of implementation of the RETE match algorithm and provides a program development framework for RETE and its extensions.