Spline Fitting through Parallel Processing.

Abstract

In many applications there appears the need to represent a set of raw data by fitting to them a smooth function or set of functions. A great amount of theoretical work has been done within the framework of approximation theory in studying properties of curve fitting for various families of approximating functions. One of the most attractive and well structured families of approximating functions are the splines. They have been extensively studied in the mathematical literature. Splines have been very useful in statistics. In the engineering literature, spline functions have been used as approximating tools, in the areas of Systems and Pattern Recognition. In this paper the authors concentrate on a fast, parallel computation technique for fitting a spline to n equispaced data points. Existing techniques can fit splines in 0(n) time, with recursive (serial) processing of the data. The new technique is based on the use of Fast Fourier Transform, and its parallel processing capability. As a result, our technique achieves the fitting of a spline in 0(log sub 2 n) time. The n data points must be equispaced.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 01, 1983
Accession Number
ADA127196

Entities

People

  • D. Kazakos

Organizations

  • University of Virginia

Tags

Communities of Interest

  • Biomedical
  • C4I
  • Space

DTIC Thesaurus Topics

  • Algorithms
  • Artificial Intelligence
  • Computations
  • Computer Science
  • Eigenvalues
  • Eigenvectors
  • Electrical Engineering
  • Engineering
  • Fast Fourier Transforms
  • Information Science
  • Integrals
  • Materials Science
  • Military Research
  • Parallel Computing
  • Parallel Processing
  • Pattern Recognition
  • Signal Processing

Readers

  • Approximation Theory.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms