Neural networks and their diagnostic applications
Chris M. Bishop · Review of Scientific Instruments · 1992
Neural network techniques offer a wide range of new opportunities for the analysis of data from plasma diagnostics. In particular, the class of neural network known as the multilayer perceptron provides a general purpose approach to nonlinear data transformation between multidimensional spaces. In this paper, we outline the principles of the multilayer perceptron, and illustrate its application to plasma diagnostics using two examples. The first of these concerns the extraction of line shape parameters from spectral data, and offers considerable improvements in speed compared with conventional approaches. The second application involves deconvolution of line integral data from a multichannel interferometer, allowing the extraction of more detailed density profiles than obtained by conventional Abel inversion.