## Artificial Neural Networks

The concept of neural networks (cf. Hertz et al. 1990) has it roots in brain research but has a counterpart in artificial intelligence, where it also involves elements of approximation theory and mathematical optimization. During the training phase, the artificial neural network is adjusted to a set of observations by a learning algorithm.

In approximation theory, this corresponds to the computations of a set of weighting coefficients associated with some basis functions. After this phase, the tuned network computes output based on further input.

Neural networks have been used by Sarro et al. (2006) to separate pulsating stars from EBs and to classify EBs into four categories according to their characteristics such as eclipse depth and widths. The classification is performed by a Bayesian ensemble of neural networks trained with HIPPARCOS data of seven different categories including eccentric binary systems and two types of pulsating light curve morphologies. In a follow-up step, these four categories are related to the configurations detached, semi-detached and over-contact binaries.

Whereas Sarro et al. (2006) use artificial neural networks for automatic classification, Devinney et al. (2006) used it in their project Eclipsing Binary Artificial Intelligence (EBAI) for deriving starting parameters. They employed a neural network to "map" observational data to approximate model elements. Observational data for an EB 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.

Prsa et al. (2008) constructed a three-layer back-propagation neural network that solves nonlinear regression problems. They describe the basic concepts and procedures for applying artificial neural networks to detached EBs. Their neural network was trained with 33,235 WD-generated light curves and applied to a set of 10,000 synthetic detached EBs, to 50 detached binaries from the Catalog and AtLas of Eclipsing Binaries (CALEB1) and to the set of 2,580 OGLE LMC binaries (Wyrzykowski et al. 2003) classified as detached.

## Telescopes Mastery

Through this ebook, you are going to learn what you will need to know all about the telescopes that can provide a fun and rewarding hobby for you and your family!

## Post a comment