VECTOR GENERATIONS IN NEURAL NETWORK COMPUTATIONS

摘要

Data pathways are important in layered neural networks.rnThe problem is how to classify information pathways inrnthe network computations. First, the architecture of thernbiological asymmetric network with odd-even (or evenodd)rnorder nonlinearities is analyzed for the networkrncomputations. The stimulus with a mixture distribution isrnuseful to evaluate their network processing ability for thernmovement direction and its velocity, which generate arnvector. Then, white noise analysis is applied to solve thernproblem. Thus, the characterized equation is derived inrnthe network computations. which evaluates the processingrnability of the network. Second, the movement velocity isrnderived, which is represented in Wiener kernels of thernnetwork computations. Thus, the information pathwaysrnfor characterizing the ability of the movement detectionrnare classified for the layered neural networksrncomputations.