随机逼近两类信息的第n个最小误差之间的关系.pdfVIP

随机逼近两类信息的第 n 个最小误差之间的关系 Abstract Multivariate approximation is one of the most popular problems in IBC (In- formation Based Complexity). It is studied commonly by a lot of mathematicians. The most important reason is that it is related to continuous problems essentially. For example, multivariate integration, solutions of partial differential equations, solu- tions of integral equations and some non-linear problems and so on. The nth minimal error is one of the most important problems in IBC. It has a lot to do with corresponding complexity of information and computation. It is the best possible algorithm among all algorithms using at most n information operations and ε−complexity is based on it. Then we can focus on the study of tractability of mul- tivariate integration and approximation which was introduced in 1994 by professor Woźniakowski from Columbia university in New York, U. S. A. Therefore, it is ne- cessary to study the nth minimal error. For the nth minimal error, there are many details in IBC, including the defini- tion of it in the worst case setting, average setting, and randomized setting, and the theories of the nth minimal error. According to the theorems of the nth minimal er- rors and the proof of some theorems, we obtain lots of evidences and methods which are good for our further study of ε− complexity and the tractability of multivariate problems. Especially, constructing bump function to get the lower bounds of the nth minimal errors and ε− complexity is an important way to solve other similar prob- lems in different settings and spaces. In this paper, two classes of information are considered. One is linear informa- tion which consists of all (continuous) linear functionals, denoted by all