Signal generator utilizing a neural network

A method for operating a data processing system to calibrate a signal generator having a plurality of signal generator input parameters, X, that produces a signal having calculated characteristics, Y, said method comprising:

• causing said data processing system to generate a plurality of sets of said plurality of signal generator input parameters, Xi;
• for each set of said plurality of signal generator input parameters, Xi, generating signals output from said signal generator in response to the set of said plurality of signal generator input parameters, Xi, and calculating a corresponding set of calculated parameters, Yi;
• training an artificial neural network using said sets of said plurality of signal generator input parameters, Xi, and said sets of calculated parameters, Yi, as a training set to train an artificial neural network to produce an inverse function that maps Y values of the sets of calculated parameters, Yi, to corresponding X values of said sets of said plurality of signal generator input parameters, Xi, wherein the X values are between predetermined limits that represent a range of possible X values for the signal generator input parameters, respectively;
• applying the trained artificial neural network to input a plurality of calculated parameters from a test signal corresponding to a plurality of signal generator input parameters and to output values for a plurality of revised signal generator input parameters by mapping the plurality of calculated parameters to the revised signal generator input parameters using the inverse function;
• and
• controlling the signal generator to generate another test signal using the revised signal generator input parameters provided by the trained artificial neural network.

A non-transitory computer readable medium containing instructions that, when executed by a computer, cause the computer to:

• generate a plurality of sets of signal generator input parameters, Xi;
• for each set of signal generator input parameters, Xi, controlling a signal generator to generate signals in response to the set of signal generator input parameters, Xi;
• calculate a plurality of sets of calculated characteristics, Yi, corresponding to the plurality of sets input parameters, Xi respectively, wherein the plurality of sets of calculated characteristics, Yi, characterize the signals;
• train an artificial neural network using said sets of signal generator input parameters, Xi, and said plurality of sets of calculated parameters, Yi, as a training set to an artificial neural network to produce an inverse function that maps Y values of the sets of calculated parameters, Yi, to corresponding X values of said sets of signal generator input parameters, Xi, wherein the X values are between predetermined limits that represent a range of possible X values for the signal generator input parameters respectively;
• apply the trained artificial neural network to input a plurality of calculated parameters from a test signal corresponding to a plurality of signal generator input parameters and to output values for a plurality of revised signal generator input parameters by mapping the plurality of calculated parameters to the revised signal generator input parameters using the inverse function;
• and
• control the signal generator to generate another test signal using the revised signal generator input parameters provided by the trained artificial neural network.