| 摘要 |
<p>PROBLEM TO BE SOLVED: To provide a digital-to-analog fitting apparatus capable of analyzing the output of a discrete time linear system in which a matrix A representing the discrete time linear system is not diagonalizable and also capable of suppressing the increase of the amount of calculation even when the number of samples, or the output of the discrete time linear system, increases.SOLUTION: A digital-to-analog fitting apparatus calculates, from a discrete time linear system containing a matrix of the Jordan normal form, a constant coefficient simultaneous heterogeneous differentiation equation, by using a monotone increase function which can be differentiated at least once and is staircase-like shaped, and constructs a neural network to which added are a weighting coefficient given to the neuron state quantity transmitted by a connection between neurons, and an added value added to each neuron. The digital-to-analog fitting apparatus operates the neural network, and obtains a fitting curve of the output of the discrete time linear system.</p> |
