An embedded connectionist approach for the inverse shortest paths problem

A type of connectionist network, called the binary relation inference network, has been recently applied to solve constrained optimization problems, such as the shortest path problem, assignment problem, etc. The inherently parallel operating nature of the network promises a potential speedup in solving the problems. In some situations where the problems cannot be solved directly with the network, it is possible to have the network acting as an embedded real-time engine to solve the involved subproblems. In this paper, the possibility of embedding the network to solve the inverse shortest paths problem is explored. Limitations in incorporating the network are discussed and remedies are suggested