ON OPTIMAL-SOLUTIONS OF THE DECONVOLUTION PROBLEM

被引：7

作者：

ERMAKOV, M

机构：

[1] Sci. Res. Inst. of Phys., Leningrad State Univ.

来源：

INVERSE PROBLEMS | 1990年 / 6卷 / 05期

关键词：

D O I：

10.1088/0266-5611/6/5/012

中图分类号：

O29 [应用数学];

学科分类号：

070104 ;

摘要：

The asymptotic behaviour of minimax and Bayes risks are studied for the deconvolution problem as a power of noise tends to zero. Under wide assumptions the author shows that the convergence order of the regularization method does not depend on the regularizer and is optimal for both Bayes and minimax risks. The author finds the connection between the orders of convergence of minimax and Bayes risks and proposes a simple algorithm for their calculation.

引用

页码：863 / 872

页数：10

共 15 条

[1]

AREFIEVA M, 1985, J COMPUT MATH MATH P, V25, P654

[2]

AREFIEVA M, 1986, J COMPUT MATH MATH P, V26, P23

[3]

ERMAKOV M, 1989, PROBL T INFORM, V25, P28

[4]

ERMAKOV M, 1986, 1ST WORLD C PROB THE, V1, P181