K. KhompurngsonB. NovaprateepD. PoltemUniversity of PhayaoPERDOMahidol UniversityBurapha University2018-06-112018-06-112012-12-01International Journal of Mathematics and Computers in Simulation. Vol.6, No.6 (2012), 513-520199801592-s2.0-84875699870https://repository.li.mahidol.ac.th/handle/20.500.14594/14027Abstract-The intend of learning problem is to identify the best predictor from given data. Specifically, the well-known hypercircle inequality was applied to kernel-based machine learning when data is know exactly. In our previous work, this lead us to extend it to circumstance for which data is known within error. In this paper, we continues the study of this subject by improving the hypothesis of nonlinear optimization problem which is used to obtain the best predictor. In additional, we apply our results to special problem of learning the value of a function from inaccurate data with different error tolerance of data error.Mahidol UniversityComputer ScienceMathematicsArticleSCOPUS