Właściwości metrologiczne radialnych i sigmoidalnych sieci neuronowych zastosowanych do korekcji błędów statycznych w przetworniku próbkującym

The paper describes the results of simulation studies of RBF and MLP neural networks. There was modeled a measurement situation in which neural networks performed static error correction of a non-linear sensor in a sampling transducer. The correction inaccuracy by neural networks is expressed by the uncertainty. There were compared metrological properties of both types of networks depending on their structures and size of learning sets.