Premium
Error performance analysis of artificial neural networks applied to Rutherford backscattering
Author(s) -
Vieira A.,
Barradas N. P.,
Jeynes C.
Publication year - 2001
Publication title -
surface and interface analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.52
H-Index - 90
eISSN - 1096-9918
pISSN - 0142-2421
DOI - 10.1002/sia.949
Subject(s) - artificial neural network , backpropagation , generality , reliability (semiconductor) , range (aeronautics) , computer science , experimental data , data set , code (set theory) , set (abstract data type) , algorithm , artificial intelligence , materials science , mathematics , statistics , physics , psychology , power (physics) , quantum mechanics , composite material , psychotherapist , programming language
We have developed a code based on artificial neural networks (ANN) to analyse Rutherford backscattering data. In particular, we have applied the code to the analysis of germanium implants in silicon substrates. Here, we study the reliability and accuracy of the quantitative results obtained. We first constructed three different training data sets. The first data set was fully general. On the second one, we restricted the experimental conditions to well‐defined values, and on the third we also restricted the implantation parameters (depth and dose of implant) to a narrower range. We then studied the trade‐off between generality and accuracy of the ANNs obtained. Furthermore, for a given architecture we applied two different training processes. The first was backpropagation on the whole data set. In the second we excluded, after an initial training phase, all the training cases with errors double the average and then continued training. Each of the processes was applied to the three different data sets. We report the performance of the ANNs so obtained when applied to real experimental data. Copyright © 2001 John Wiley & Sons, Ltd.