Analytic formulation of intelligent machines as neural nets

Abstract

The design of the organization level of the intelligent machine as a Boltzmann machine, as described in current neural network literature, is discussed. Since this level is responsible for planning the actions of the machine, the problem at this tier is formulated as the construction of the right sequence of tasks or events which minimizes the entropy for the desired action. Two search techniques, simulated annealing (SA) and expanding subinterval random search (ESRS), are described. These techniques are used to find the global minimum entropy of a Boltzmann machine. Simulations using these search techniques were conducted using energy as a cost function, and results indicate that ESRS converges faster than SA to a global minimum if the topology contains narrow and deep cost wells.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>