NEURAL NETWORK TECHNIQUE USED FOR MODELING NONLINEAR ERROR DATA.Researchers at NIST (National Institute of Standards & Technology, Washington, DC, www.nist.gov) The standards-defining agency of the U.S. government, formerly the National Bureau of Standards. It is one of three agencies that fall under the Technology Administration (www.technology. have developed a new technique for modeling nonlinear error data that can reduce the amount of testing that a linear model would require. This technique can also be used to predict the errors over a full set of test points, based on only the selected subset of test points. It is expected that this nonlinear modeling approach can be incorporated into the NIST High-dimensional Empirical Linear Prediction Linear prediction is a mathematical operation where future values of a discrete-time signal are estimated as a linear function of previous samples. In digital signal processing, linear prediction is often called linear predictive coding (LPC) and can thus be viewed as a (HELP) Toolbox, which will soon be available on the NIST Web site. A Neural Network (NN) Toolbox was used in conjunction with second-order Gradient Search software to build five-layer NN models of simulated nonlinear data. The limitations of this approach were explored in terms of speed of convergence, noise immunity, and level of complexity. In modeling a five-parameter bandpass filter, for example, it was found that a five-fold improvement in the mean squared error In statistics, the mean squared error or MSE of an estimator is the expected value of the square of the "error." The error is the amount by which the estimator differs from the quantity to be estimated. was obtained using a neural network nonlinear modeling approach versus a linear model when the parameter values of the filter were changed by 20%. Other examples of interest were examined for this purpose, including the data from a thermal-based voltmeter. Using a 45-parameter model, the NN modeling showed significantly better results than linear modeling when the number of test points was less than 62. A paper describing this research was presented at the IEEE (Institute of Electrical and Electronics Engineers, New York, www.ieee.org) A membership organization that includes engineers, scientists and students in electronics and allied fields. Instrumentation and Measurement Technology Conference (IMTC (International Multimedia Telecommunications Consortium, San Ramon, CA, www.imtc.org) An international membership organization founded in 1993 as Consortium for Audiographics Teleconferencing Standards (CATS). ) 2000 held in Baltimore, MD. |
|
||||||||||||||||
Printer friendly
Cite/link
Email
Feedback
Reader Opinion