engleski

Self-Organizing Polynomial Networks for Time-Constrained Applications

A meta-modeling of complex calculation procedures (static systems) is investigated with the aim to create the low complexity surrogate models applicable in low computing power real-time measurement systems. Unlike the single parameter error criteria and MDL measure, the proposed compound squared relative error measure forces the GMDH algorithm to prefer the models having the smallest compound deviation of the accuracy and the execution time from the given thresholds and thus generally leads to more favorable models with respect to both conditions. The approximation errors, the execution speed and the applicability of derived GMDH models in real-time flowrate measurements of natural gas are discussed and compared with the corresponding models derived by ANN and SVR.

GMDH ; approximation methods ; computational intelligence ; measurement uncertainty ; natural gas ; polynomials ; training ; approximation methods ; embedded systems ; modeling ; real-time systems

nije evidentirano