This paper presents learning rates for the least-square regularized regression algorithms with polynomial kernels. The target is the error analysis for the regression problem in learning theory. A regularization schem...This paper presents learning rates for the least-square regularized regression algorithms with polynomial kernels. The target is the error analysis for the regression problem in learning theory. A regularization scheme is given, which yields sharp learning rates. The rates depend on the dimension of polynomial space and polynomial reproducing kernel Hilbert space measured by covering numbers. Meanwhile, we also establish the direct approximation theorem by Bernstein-Durrmeyer operators in $ L_{\rho _X }^2 $ with Borel probability measure.展开更多
Semi-supervised learning has been of growing interest over the past few years and many methods have been proposed. Although various algorithms are provided to implement semi-supervised learning,there are still gaps in...Semi-supervised learning has been of growing interest over the past few years and many methods have been proposed. Although various algorithms are provided to implement semi-supervised learning,there are still gaps in our understanding of the dependence of generalization error on the numbers of labeled and unlabeled data. In this paper,we consider a graph-based semi-supervised classification algorithm and establish its generalization error bounds. Our results show the close relations between the generalization performance and the structural invariants of data graph.展开更多
The concept of computability is defined more exactly and illustrated as an example of Boolean functions and cryptanalysis. To define a Boolean function is not necessary to record its formula. To do that the reduced (c...The concept of computability is defined more exactly and illustrated as an example of Boolean functions and cryptanalysis. To define a Boolean function is not necessary to record its formula. To do that the reduced (compact) description of values is determined in the truth table or in the statement of the problem. We obtain estimates of computation time, the volume of a compact descriptions and the range of variables under which it takes the value 0 or 1, depending polynomially on the number of arguments.展开更多
Corrected explicit-implicit domain decomposition(CEIDD) algorithms are studied for parallel approximation of semilinear parabolic problems on distributed memory processors. It is natural to divide the spatial domain i...Corrected explicit-implicit domain decomposition(CEIDD) algorithms are studied for parallel approximation of semilinear parabolic problems on distributed memory processors. It is natural to divide the spatial domain into some smaller parallel strips and cells using the simplest straightline interface(SI) . By using the Leray-Schauder fixed-point theorem and the discrete energy method,it is shown that the resulting CEIDD-SI algorithm is uniquely solvable,unconditionally stable and convergent. The CEIDD-SI method always suffers from the globalization of data communication when interior boundaries cross into each other inside the domain. To overcome this disadvantage,a composite interface(CI) that consists of straight segments and zigzag fractions is suggested. The corresponding CEIDD-CI algorithm is proven to be solvable,stable and convergent. Numerical experiments are presented to support the theoretical results.展开更多
文摘This paper presents learning rates for the least-square regularized regression algorithms with polynomial kernels. The target is the error analysis for the regression problem in learning theory. A regularization scheme is given, which yields sharp learning rates. The rates depend on the dimension of polynomial space and polynomial reproducing kernel Hilbert space measured by covering numbers. Meanwhile, we also establish the direct approximation theorem by Bernstein-Durrmeyer operators in $ L_{\rho _X }^2 $ with Borel probability measure.
基金supported by National Natural Science Foundation of China (Grant No. 10771053)Specialized Research Foundation for the Doctoral Program of Higher Education of China (SRFDP) (Grant No. 20060512001)Natural Science Foundation of Hubei Province (Grant No. 2007ABA139)
文摘Semi-supervised learning has been of growing interest over the past few years and many methods have been proposed. Although various algorithms are provided to implement semi-supervised learning,there are still gaps in our understanding of the dependence of generalization error on the numbers of labeled and unlabeled data. In this paper,we consider a graph-based semi-supervised classification algorithm and establish its generalization error bounds. Our results show the close relations between the generalization performance and the structural invariants of data graph.
文摘The concept of computability is defined more exactly and illustrated as an example of Boolean functions and cryptanalysis. To define a Boolean function is not necessary to record its formula. To do that the reduced (compact) description of values is determined in the truth table or in the statement of the problem. We obtain estimates of computation time, the volume of a compact descriptions and the range of variables under which it takes the value 0 or 1, depending polynomially on the number of arguments.
基金supported by National Natural Science Foundation of China (Grant No. 10871044)
文摘Corrected explicit-implicit domain decomposition(CEIDD) algorithms are studied for parallel approximation of semilinear parabolic problems on distributed memory processors. It is natural to divide the spatial domain into some smaller parallel strips and cells using the simplest straightline interface(SI) . By using the Leray-Schauder fixed-point theorem and the discrete energy method,it is shown that the resulting CEIDD-SI algorithm is uniquely solvable,unconditionally stable and convergent. The CEIDD-SI method always suffers from the globalization of data communication when interior boundaries cross into each other inside the domain. To overcome this disadvantage,a composite interface(CI) that consists of straight segments and zigzag fractions is suggested. The corresponding CEIDD-CI algorithm is proven to be solvable,stable and convergent. Numerical experiments are presented to support the theoretical results.
