To have a general case study of the proposed algorithm for
general optimization of modular neural networks for person recognition,
we initially considered a benchmark database of human iris
images. Three types of learning algorithms (gradient descent with
adaptive learning (GDA), gradient descent with adaptive learning
and momentum (GDX) and scaled conjugate gradient (SCG)) are
used for testing with different sizes of the cut in iris images. In analyzing
the results in each module we arrived to the conclusion that