The grand tour: a tool for viewing multidimensional data
SIAM Journal on Scientific and Statistical Computing
The elements of graphing data
A Bayesian approach to on-line learning
On-line learning in neural networks
Reconstruction and representation of 3D objects with radial basis functions
Proceedings of the 28th annual conference on Computer graphics and interactive techniques
Radial Basis Functions
HyperSlice: visualization of scalar functions of many variables
VIS '93 Proceedings of the 4th conference on Visualization '93
Comparing Dot and Landscape Spatializations for Visual Memory Differences
IEEE Transactions on Visualization and Computer Graphics
Programming Massively Parallel Processors: A Hands-on Approach
Programming Massively Parallel Processors: A Hands-on Approach
High Performance Matrix Inversion on a Multi-core Platform with Several GPUs
PDP '11 Proceedings of the 2011 19th International Euromicro Conference on Parallel, Distributed and Network-Based Processing
IEEE Transactions on Visualization and Computer Graphics
Computational steering for patient-specific implant planning in orthopedics
EG VCBM'08 Proceedings of the First Eurographics conference on Visual Computing for Biomedicine
Hypermoval: interactive visual validation of regression models for real-time simulation
EuroVis'10 Proceedings of the 12th Eurographics / IEEE - VGTC conference on Visualization
Uncertainty-aware exploration of continuous parameter spaces using multivariate prediction
EuroVis'11 Proceedings of the 13th Eurographics / IEEE - VGTC conference on Visualization
Hi-index | 0.00 |
In this paper we present a technique which allows us to perform high quality and progressive response surface prediction from multidimensional input samples in an efficient manner. We utilize kriging interpolation to estimate a response surface which minimizes the expectation value and variance of the prediction error. High computational efficiency is achieved by employing parallel matrix and vector operations on the GPU. Our approach differs from previous kriging approaches in that it uses a novel progressive updating scheme for new samples based on blockwise matrix inversion. In this way we can handle very large sample sets to which new samples are continually added. Furthermore, we can monitor the incremental evolution of the surface, providing a means to early terminate the computation when no significant changes have occurred. When the generation of input samples is fast enough, our technique enables steering this generation process interactively to find relevant dependency relations.