Approximation by fuzzy B-spline series (Q815981)

For the fuzzy B-spline series \(S(x)=\sum_{j\in{\Lambda}}\) \(N_{j}(x)\cdot{C_j}\) on \(A\) \((A=\mathbb{R}\) or \(A=[a,b])\) with coefficients \(C_j\) of fuzzy numbers and crisp B-splines \(N_j(x)=N_j(x;t_j,\dots,{t_{j+r}})\), where \(t_{-r+1},\dots,t_{n+r}\) are knots in \(\mathbb{R}\) \((\Lambda =\mathbb{Z}\text{ or }\Lambda =\{-r+1,\cdots,n\})\), the authors prove that \(S(x)\) is continuous as the fuzzy-number-valued function of \(x\), the knots and the coefficients, \(S(x)\) changes its sign at most as many times as the sequence \((C_j)_{j\in{\Lambda}}\) changes its sign and the length of the support of \(S(x)\) does not exceed the maximum length of the support of the fuzzy numbers \(C_j\). The paper also obtains the error bounds for the approximation of a continuous fuzzy-number-valued function by fuzzy B-spline series in terms of the modulus of continuity and the particularly simple error bounds for fuzzy B-spline series of Schoenberg type. Finally, the authors provide two numerical examples and some further research topics.