1991 Volume 2 Pages 53
Information or embedded patterns can be extracted by via minimum energy techniques or maximum entropy methods. We show how to combine these two approaches via Bayes' theorem so that the problem can then be cast into that of minimizing the Helmholtz free energy. This problem can then be solved an optimization algorithm such as simulated annealing which as we show yields the minimum Helmholtz free energy state rather than the minimum energy state. We will also briefly discuss the Bayes'theorem as a model for learning.