Sign on

SAO/NASA ADS Astronomy Abstract Service


· Find Similar Abstracts (with default settings below)
· Full Refereed Journal Article (PDF/Postscript)
· Full Refereed Scanned Article (GIF)
· References in the article
· Citations to the Article (25) (Citation History)
· Refereed Citations to the Article
· Reads History
·
· Translate This Page
Title:
Optimization algorithms - Simulated annealing and neural network processing
Authors:
Jeffrey, W.; Rosner, R.
Affiliation:
AA(Harvard-Smithsonian Center for Astrophysics, Cambridge, MA)
Publication:
Astrophysical Journal, Part 1 (ISSN 0004- 637X), vol. 310, Nov. 1, 1986, p. 473-481. (ApJ Homepage)
Publication Date:
11/1986
Category:
NUMERICAL ANALYSIS
Origin:
STI
NASA/STI Keywords:
Algorithms, Neural Nets, Optimization, Simulated Annealing, Superconducting Devices, Systems Simulation, Annealing, Extremum Values, Fredholm Equations, Ill-Posed Problems (Mathematics), Remote Sensing
DOI:
10.1086/164700
Bibliographic Code:
1986ApJ...310..473J

Abstract

Two algorithms which have been previously used for discrete optimization, simulated annealing and neural network processing, are developed and compared. It is demonstrated how these algorithms can be used to find global extrema of functions, while avoiding trapping in local extrema. In the standard treatment of neural network processors, only quadratic and linear terms in the function variables are included in the objective function. This traditional approach is extended to show how constraints not expressible in quadratic and linear terms (e.g., entropy) can be incorporated into the function to be minimized. The efficiency of the implementation of neural net processing is also demonstrated, and it is shown how its speed advantage over more traditional optimization techniques (even when implemented on serial processors) is related to its convergence properties. An important application of the results is in the interpretation of remote sensing data, since typical indirect sensing problems can be readily cast into the language of optimization theory; the methods presented here have the particular ability to solve severely ill posed inversion problems. The algorithms described here have been implemented on a serial processor but are cast in a form which is ideally suited for parallel processing.

Printing Options

Print whole paper
Print Page(s) through

Return 600 dpi PDF to Acrobat/Browser. Different resolutions (200 or 600 dpi), formats (Postscript, PDF, etc), page sizes (US Letter, European A4, etc), and compression (gzip,compress,none) can be set through the Printing Preferences



More Article Retrieval Options

HELP for Article Retrieval


Bibtex entry for this abstract   Preferred format for this abstract (see Preferences)


Find Similar Abstracts:

Use: Authors
Title
Keywords (in text query field)
Abstract Text
Return: Query Results Return    items starting with number
Query Form
Database: Astronomy
Physics
arXiv e-prints