Model calculation unit and controller for selectively calculating an RBF model, a Gaussian process model and an MLP model

摘要

The invention relates to a model calculation unit (22) for selectively calculating an RBF model or a neuron layer of a multilayer perceptron model with a hardwired computation core (11, 13, 14) designed to calculate a fixed computation algorithm in coupled function blocks, wherein the computational kernel ( 11, 13, 14) is designed to produce an output variable (y) for an RBF model depending on one or more input variables of an input variable vector (ut), interpolation point points (V), length scales (L) and parameters specified for each interpolation point point. wherein the arithmetic kernel (11, 13, 14) is further adapted to provide, for the neuron layer of the multi-layered perceptron model, a number of neurons (20) dependent on the one or more input variables of the input vector (u) of a weighting matrix Weighting factors (v0 ... p7-1,0 ... p6-1) and one for each s neuron (20) predetermined offset value to calculate an output variable for each neuron (20).