Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Cardinal interpolation.

Steven C Gustafson1, David R Parker, Richard K Martin

  • 1Air Force Institute of Technology, Wright Patterson AFB, OH 45433, USA. steven.gustafson@afit.edu

IEEE Transactions on Pattern Analysis and Machine Intelligence
|July 14, 2007
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Recoded Bacteriophage Genome for Bio-Orthogonal-Enabled Concentration and Detection of <i>E. coli</i> in Drinking Water.

ACS synthetic biology·2025
Same author

Pixel material classification using polarimetric LiDAR active imaging.

Applied optics·2025
Same author

Bacteriophage-Based Bioanalysis.

Annual review of analytical chemistry (Palo Alto, Calif.)·2024
Same author

Simulation of anisoplanatic lucky look imaging and statistics through optical turbulence using numerical wave propagation: erratum.

Applied optics·2022
Same author

Atmospheric optical turbulence mitigation using iterative image registration and least squares lucky look fusion.

Applied optics·2022
Same author

Simulation of anisoplanatic lucky look imaging and statistics through optical turbulence using numerical wave propagation.

Applied optics·2021

A new Bayesian probability density enables cardinal interpolation, ensuring accurate extrapolation for interpolating functions. This method provides a smooth mean and variance, extending to least squares linear models.

Area of Science:

  • Computational Statistics
  • Machine Learning
  • Numerical Analysis

Background:

  • Interpolating functions are crucial in data analysis but often lack reliable extrapolation properties.
  • Existing methods struggle to provide a probabilistic framework that guarantees extrapolation to simpler models like linear regression.
  • The need for a density that behaves predictably beyond the observed data points is a significant challenge.

Purpose of the Study:

  • To develop a Bayesian probability density for interpolating functions with a novel property called cardinal interpolation.
  • To demonstrate the desirable properties and practical potential of this new density, particularly its extrapolation capabilities.
  • To extend this extrapolation property to more general approximating functions beyond strict interpolation.

Main Methods:

Related Experiment Videos

  • Development of a Bayesian probability density function for interpolating functions.
  • Application of fully Bayesian methods to Gaussian radial basis interpolators.
  • Optimization of interpolator smoothness to determine basis function widths.
  • Mathematical analysis to derive the properties of the mean and variance of the interpolating density.

Main Results:

  • Introduction of cardinal interpolation, a property ensuring extrapolation to the least squares linear model density.
  • The mean of the cardinal interpolation density smoothly interpolates data points and extrapolates to the least squares line.
  • The variance is zero at data points, increases with distance, and extrapolates to the quadratic variance of the least squares line.
  • The resulting density is non-Gaussian, optimizing smoothness and not arising from a standard Gaussian process.

Conclusions:

  • The developed Bayesian cardinal interpolation density offers a robust solution for data modeling with reliable extrapolation.
  • This method provides a principled way to connect complex interpolations with simpler, well-understood linear models.
  • The framework is extendable to general approximating functions, broadening its applicability in statistical modeling and machine learning.