Computer Science – Artificial Intelligence
Scientific paper
Jun 2006
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2006aas...208.4409d&link_type=abstract
American Astronomical Society Meeting 208, #44.09; Bulletin of the American Astronomical Society, Vol. 38, p.119
Computer Science
Artificial Intelligence
5
Scientific paper
Many problems in astronomy require solution of inverse problems. Nowadays we are increasingly able to create models of phenomena, wherein our challenge is to determine the best parameters that match observations for which the model is valid. We often determine these parameters via a differential corrections procedure, which requires starting point parameters. If these are not close to the solution, the differential corrections procedure is liable to not converge.Formerly the discovery rate of objects requiring solution using models allowed for human-mediated guessing of initial parameters. However, discovery rates, for example of variable stars including eclipsing binaries, are increasing rapidly as more powerful ground-based telescopes and new satellites come on line, and we could expect discoveries in the millions. To extract astrophysical information from even a small fraction of these, humans cannot be in the data-reduction loop and we are developing techniques using artificial intelligence to eliminate this requirement. We are developing an “Intelligent Data Pipeline” architecture in which an eclipsing binary solution is divided in two parts: a module for obtaining a starting solution; and a differential corrections module.In one approach for a starting solution we employ a neural network to “map” observational data to approximate model elements. Observational data for an eclipsing binary are presented to the network’s input nodes and its output nodes yield starting model elements. The network is first “trained” on many observational data-model element pairs. We are also developing a second and perhaps novel approach employing models computed over a grid of parameters and stored, plus a means to index the appropriate model elements for observations of a particular eclipsing binary. The method appears applicable to any domain where models are confronted by observations.This research is supported by NSF/RUI Grant AST05-07542.
Bradstreet David H.
DeGeorge M.
Devinney Edward J. Jr.
Giammarco J. M.
Guinan Edward
No associations
LandOfFree
Observe the Model: On Inverse Problems in Astronomy does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Observe the Model: On Inverse Problems in Astronomy, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Observe the Model: On Inverse Problems in Astronomy will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1856591