Computer Science – Performance
Scientific paper
Dec 2005
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2005agufmsm11a..04o&link_type=abstract
American Geophysical Union, Fall Meeting 2005, abstract #SM11A-04
Computer Science
Performance
7924 Forecasting (2722), 7954 Magnetic Storms (2788), 7959 Models
Scientific paper
There now exists a variety of empirical models of three magnetic storm phenomena: The ring current index Dst, >2 MeV electron flux at geosynchronous orbit, and the location of the plasmapause. These three phenomena are connected by the fact that each one exhibits nearly all of its dynamics during magnetic storms. Within each magnetic storm the magnetospheric drivers and responses are highly correlated with each other and with themselves. The episodic and autocorrelated nature of the dynamics leads to a challenge when building empirical models: each storm essentially counts as one independent sample of the dynamic process one is trying to model. Thus, each year presents only about 30 storms, or 30 samples. Any method for constructing an empirical model that adheres strictly to the rule of parsimony (Occam's Razor) will result in a minimum required correlation coefficient for a given level of complexity for a given size of the data set used to optimize the model. In this paper we estimate the minimum required correlation coefficient versus the number of free parameters in a broad class of empirical models. We conclude that empirical models of magnetospheric phenomena require either longer data sets for optimization or inputs that are better correlated with the phenomenon being modeled. Conversely, greater flexibility in empirical modeling, which nearly always demands greater complexity, seems unlikely to yield substantial improvements in the quantitative performance of the models.
No associations
LandOfFree
Limits on the Complexity of Empirical Models of Magnetic Storm Phenomena 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 Limits on the Complexity of Empirical Models of Magnetic Storm Phenomena, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Limits on the Complexity of Empirical Models of Magnetic Storm Phenomena will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-759741