Forecasting chaotic cardiovascular time series with an adaptive slope multilayer perceptron neural network

A multilayer perceptron (MLP)network architecture has been formulated in wh ich two adaptive parameters, the scaling and translation of the postsynapti c function at each node, are allowed to adjust iteratively by gradient-desc ent. The algorithm has been employed to predict experimental cardiovascular time series, following systematic reconstruction of the strange attractor of the training signal, Comparison with a standard MLP employing identical numbers of nodes and weight learning rates demonstrates that the adaptive a pproach provides an efficient modification of the MLP that permits faster l earning. Thus, for an equivalent number of training epochs there was improv ed accuracy and generalization for both one- and L-step ahead prediction. T he applicability of the methodology is demonstrated for a set of monotonic postsynaptic functions (sigmoidal, upper bounded, and nonbounded), The appr oach is computationally inexpensive as the increase in the parameter space of the network compared to a standard MLP is small.