Paper Title
Mutual Information Based Ensemble Support Vector Data Description

Abstract
Hyperspectral imagery (HSI) has spatial and detailed spectral information. Therefore, it has been used in many different areas for anomaly and target detection or classification problems. Because it consists large amount of data, effective, accurate and fast computational methods have become critical issue in machine learning. The support vector data description (SVDD) is one of the powerful methods as a one-class classifier for classification problems in machine learning area. It is a non-parametric boundary method that tries to enclose the target objects in a minimum hypersphere as much as possible. Using kernel function is one of its advantages. The kernelization makes SVDD more efficient algorithms, when the objective data is not spherically distributed. Apart from using kernel function, ensemble methods can be also used to improve classification performance of the SVDD. Giving proper weight to each classifier before combination is one of the important part of ensemble methods. In this paper, we have offered Mutual Information (MI) between each classifier in order to use as coefficients to weighted combinators. Keywords- Bagging; Classification; Hyperspectral imagery; Machine learning; Mutual Information; Support vector data description.