Control of a subsonic electropneumatic acoustic generator with dynamic recurrent neural networks

摘要

A dynamic recurrent neural network model is discussed, which presents two types of adaptive parameters : the classical weights between the units and the time constant associated with each artificial neuron. This dynamic neural model with recurrent connections can deal with time-varying input and output tasks in nontrivial ways. It is successfully used for linearising a subsonic electropneumatic acoustic generator, that is demonstrated to be an efficient but nonlinear acoustic generator.