Learning by Choice of Internal Representations: An Energy Minimization Approach

Learning by choice of internal representations (CHIR) is a learning algorithm for a multilayer neural network system, suggested by Grossman et al, [1,2) and based upon determining the internal representations of the system as well as its internal weights. In this paper, we propose an energy minimization approach whereby the internal representations (IR) as well as the weight matrix are allowed to be modified. Carrying out the analysis, consistency with the back propagation (BP) method (3) is demonstrated when a continuous valued system is considered, while a generalization of the CHIR learning procedure is obtained for the discrete case. Computer simulations show consistency with the results obtained by Grossman et al. for the restricted cases of parity, symmetry, and parity-symmetry problems .