Neural network realizations of Bayes decision rules for exponentially distributed data

For general Bayes decision rules there are considered perceptron approximat ions based on sufficient statistics inputs. A particular attention is paid to Bayes discrimination and classification. In the case of exponentially di stributed data with known model it is shown that a perceptron with one hidd en layer is sufficient and the learning is restricted to synaptic weights o f the output neuron. If only the dimension of the exponential model is know n, then the number of hidden layers will increase by one and also the synap tic weights of neurons from both hidden layers have to be learned.