A MEAN FIELD NEURAL NETWORK MECHANIZING COMPLEX CAUSAL REASONING PROBLEMS

A neural model for causal reasoning (also called abduction) is proposed in the open, independent, and incompatibility classes. The reasoning process is characterized in these classes by an explicit energy/target function. Potts spin mean field theory annealing methods are used to derive the mechanics of this model. Application of the model to an actual case in legal reasoning and an experimental-based comparison with a recent proposal reveal its efficiency and swiftness in computing the best solutions.