A neuron model with fluid properties for solving labyrinthian puzzle (Q756440)

The continuous Hopfield model with constraint on the weights \(T_{ii}\neq 0\) replaced by \(T_{ii}=-\Sigma T_{ij}\) for all \(j\neq i\), allows to interpret the net as a model of flowing of a liquid affected only by gravity. The potentials of neurons then ``flow'' from higher to lower levels converging to a unique stable state. An example shows how this net can be used to solve an AI problem: in a labyrinth represented by the net with connections to neighbors either blocked or free, the stable state indicates the path through the labrinth by following decreasing potentials of representing neurons.