HR: 1340h
AN: SM13A-0310 [Abstracts]
TI: Advanced Data Mining and Reverse Engineering Algorithms for Space Sciences
AU: * Valera-Guallar, L
SM13A-0310
AF: BW Analytics, 12340 El Camino Real, Suite 350, San Diego, CA 92130
United States
AU: Karimabadi, H
EM: homak@sciberquest.com
AF: SciberQuest, Inc., Pacific Executive Plaza
777 South Highway 101, Suite 108, Solana Beach, CA 92075-2623
United States
AU: White, H
EM: hal.white@bateswhite.com
AF: BW Analytics, 12340 El Camino Real, Suite 350, San Diego, CA 92130
United States
AU: Burns, P
EM: patrick.burns@bwanalytics.com
AF: BW Analytics, 12340 El Camino Real, Suite 350, San Diego, CA 92130
United States
AB:
Progress in space physics has always been strongly dependent on analysis of in situ spacecraft measurements. However, the
vast majority of spacecraft data go unexplored as most scientists have had to rely on visual inspection of the data as the
main means of mining the data. With the upcoming multi-spacecraft NASA missions (THEMIS, MMS, etc.) the growing size of data
promises to outpace the ability of scientists to analyze them. Here we present a new technique, called Relevant Input
Processor Network (RIPNet) for space plasma applications. Our recent application of this technique to modeling of
magnetopause has demonstrated superior performance metrics (speed, accuracy, etc.) compared to standard techniques. RIPNet
also offers reverse engineering capability. By this we mean that the outcome of the algorithm (i.e., the predicted model) is
an analytical function with proper dependencies on the input parameters rather than say a set of neural net connection as in
artificial neural net. This is very useful as it allows easy dissemination of results to others. The examination of the
equation can also yield information about the underlying physics and relative importance of various terms. We will
illustrate this technique through several examples.
DE: 0520 Data analysis: algorithms and implementation
DE: 0555 Neural networks, fuzzy logic, machine learning
DE: 2724 Magnetopause and boundary layers
DE: 7924 Forecasting (2722)
SC: SPA-Magnetospheric Physics [SM]
MN: Fall Meeting 2005