Stochastic finite elements: a spectral approach
Stochastic finite elements: a spectral approach
Communications of the ACM
STOC '01 Proceedings of the thirty-third annual ACM symposium on Theory of computing
Data Mining for Scientific and Engineering Applications
Data Mining for Scientific and Engineering Applications
Incremental Singular Value Decomposition of Uncertain Data with Missing Values
ECCV '02 Proceedings of the 7th European Conference on Computer Vision-Part I
Evaluating collaborative filtering recommender systems
ACM Transactions on Information Systems (TOIS)
IEEE Transactions on Knowledge and Data Engineering
Introduction to Information Retrieval
Introduction to Information Retrieval
Generalized Collocation Methods: Solutions to Nonlinear Problems
Generalized Collocation Methods: Solutions to Nonlinear Problems
| Hi-index | 0.00 |
We propose a scheme for improving existing tools for recovering and predicting decisions based on singular value decomposition. Our main contribution is an investigation of advantages of using a functional, rather than linear approximation of the response of an unknown, complicated model. A significant attractive feature of the method is the demonstrated ability to make predictions based on a highly filtered data set. An adaptive high-order interpolation is constructed, that estimates the relative probability of each possible decision. The method uses a flexible nonlinear basis, capable of utilizing all the available information. We demonstrate that the prediction can be based on a very small fraction of the training set. The suggested approach is relevant in the general field of manifold learning, as a tool for approximating the response of the models based on many parameters. Our experiments show that the approach is at least competitive with other latent factor prediction methods, and that the precision of prediction grows with the increase in the order of the polynomial basis.