Izvorni znanstveni članak

ANN cum mathematical modelling of measurement data

Lavit Rawtani ; Department of Computer Science & Electronics, Maulana Azad College of Technology, INDIA
J.L. Rana ; Department of Computer Science & Electronics, Maulana Azad College of Technology, INDIA
A.K. Tiwari ; Department of Computer Science & Electronics, Maulana Azad College of Technology, INDIA

Sažetak

Subject knowledge is often available in the form of data sets, obtained through experiments, with the possibility of some measurement errors. Mathematical modeling of such data is convenient for the use in design. ANN representation of subject knowledge is necessary for including it in the knowledge base of connectionist expert systems which are convenient for decision making. A modeling method is developed which serves as a mathematical as well as an ANN representation, combining the advantages of both. The comparison of the method is carried out with conventional methods like polynomials of different degrees, Hermite cubics with a different number of break points and cubic splines with a different number of knots. For the same error of approximation, the method is found to require a smaller number of free parameters (degrees of freedom).

Ključne riječi

ANN representation, mathematical modelling, knowledge base, expert systems, measurement data

Hrčak ID:

318832

URI

https://hrcak.srce.hr/318832

Datum izdavanja:

7.12.2000.

Posjeta: 226 *