On fitting dielectric spectra using artificial neural networks

• Published in: Bioelectrochemistry and Bioenergetics 1992-10, Vol.28 (3), p.425-434
• Main Authors: Kell, Douglas B., Davey, Christopher L.
• Format: Article
• Language: English
• Subjects: Biological and medical sciences; Fundamental and applied biological sciences. Psychology; General aspects. Models. Methods; Vertebrates: nervous system and sense organs
• ISSN: 0302-4598
• DOI: 10.1016/0302-4598(92)80030-K

In biological dielectric spectroscopy, where dispersions are substantially broader than that expected from a purely Debye-like process, it is not always possible, because of technical limitations, to obtain data over a wide enough range of frequencies to encompass the entire dispersion(s) of interest. Similarly, because of the breadth of the dispersions, it is common to seek to characterize the dielectric behaviour of interest by means of the Cole-Cole function. Whilst it is possible to fit dielectric data to this equation using appropriate non-linear least-squares methods, these methods are computationally rather demanding, and must be performed iteratively for each set of data. We show here, for the first time, that it is possible to train an artificial neural network to learn to extract the parameters of the Cole-Cole equation from small sets of dielectric data (permittivities measured at various frequencies) which can thus give an essential instantaneous output of the limiting permittivities at frequencies that are both high and low with respect to the characteristic frequency.