Massively Parallel Reasoning about Actions

In 2] C. Baral and M. Gelfond present the language A C for representing concurrent actions in dynamic systems, and give a sound but incomplete encoding of this language in terms of extended logic programming. Using their program, the time the computation of the transition from one situation to another takes increases quadraticly with the size of the considered domain. In this paper, we present a mapping of domain descriptions in A C into neural networks of linear size. These networks take only four steps for the sound and complete computation of the transition of situations. This allows for reasoning about concurrent actions wrt. extensive real-time domains.