Information processing apparatus, control method, and recording medium

摘要

An information processing apparatus selects a plurality of interval width sequences of internal knots of a nonuniform B-spline that approximates waveform data, based on a multivariate normal distribution, and then determines a knot vector based on each of the selected interval width sequences and an optimal control point sequence for each knot vector. Also, the apparatus repeats, for a nonuniform B-spline that is defined for each combination of a knot vector and an optimal control point sequence corresponding to the knot vector, the selection of interval width sequences, the determination of knot vectors and optimal control point sequences and the evaluation thereof, until an evaluation whose objective function is minimized is obtained.

主权项

1. An information processing apparatus for optimizing an arrangement of internal knots and control points of a nonuniform B-spline that approximates waveform data, comprising:
a selector which is able to select a plurality of nonuniform interval width sequences for a knot interval between internal knots, based on a multivariate normal distribution; a generator which is able to generate, for each of the plurality of interval width sequences selected by the selector, a knot vector that is based on the interval width sequence; a determiner which is able to determine an optimal control point sequence for each knot vector generated by the generator; an evaluator which is able to use an objective function to evaluate an approximation error with the waveform data, for a nonuniform B-spline defined for each combination of a knot vector and an optimal control point sequence corresponding to the knot vector; and a controller which is able to perform control so as to repeat operations including the selection of a plurality of interval width sequences by the selector, the generation of knot vectors by the generator, the determination of optimal control point sequences by the determiner, and the evaluation by the evaluator, until an evaluation indicating that the approximation error has been converged to a value regarded as a minimum is obtained by the evaluator, wherein the controller, in the case where the evaluation indicating that the approximation error has been converged to the value regarded as the minimum is not obtained by the evaluator, updates a mean and a variance of the multivariate normal distribution, based on a predetermined number of interval width sequences selected in ascending order of approximation errors from the plurality of interval width sequences selected by the selector, and causes the operations to be repeated using the multivariate normal distribution after updating.