Standard Bidirectional Associative Memory (BAM) Stores sum-of-thecorrelation-matrices of the pairs of patterns. When a pattern of an encoded pair is presented, the other is expected to be recalled. It has been shown that standard BAM cannot correctly recall a pattern pair if it is not at local minima of the energy function. To overcome this problem, novel niethods for encoding have been proposed. The efficient novel-encoding methods require knowledge of the interference noise in the standard BAM. In this paper, we propose an algorithm for computing the exact amount of interference noise in standard encoding of BAM. The computational cornplexity of the algorithm is the same as that of computing the correlation matrix for the standard BAM.
This work concentrates on a novel method for empirical estimation
of generalization ability of neural networks. Given a set of training (and testing) data, one can choose a network architecture (nurnber of layers, number of neurons in each layer etc.), an initialization method, and a learning algorithrn to obtain a network. One measure of the performance of a trained network is how dosely its actual output approximates the desired output for an input that it has never seen before. Current methods provide a “number” that indicates the estimation of the generalization ability of the network. However, this number provides no further inforrnation to understand the contributing factors when the generalization ability is not very good. The method proposed uses a number of parameters to define the generalization ability. A set of the values of these parameters provide an estimate of the generalization ability. In addition, the value of each pararneter indicates the contribution of such factors as network architecture, initialization method, training data set, etc. Furthermore, a method has been developed to verify the validity of the estimated values of the parameters.