On the Combination of Logistic Regression and Local Probability Estimates

Abstract

Abstract-In this paper we give a survey of the combination of classifiers. We briefly describe basic principles of machine learning and the problem of classifier construction and review several approaches to generate different classifiers as well as established methods to combine different classifiers. Then, we introduce our novel approach to assess the appropriateness of different classifiers based on their characteristics for each test point individually.

Pseudo Bayesian and Linear Regression Global Thresholding

“…The classifier performance depends greatly on the characteristics of the data. To take advantage of the strengths of both methods, Osl et al (2008) [19] propose an algorithm that combines a logistic regression model with a non-parametric classification method, the k-nearest neighbors. In this article we present probability models for classification problems in image analysis.…”