Classification by Boosting Differences in Input Vectors: An application to datasets from Astronomy

TL;DR

This paper introduces the Difference Boosting Neural Network (DBNN), a method that enhances differences in feature vectors to improve classification in astronomy, demonstrated on transient detection and spectral identification tasks.

Contribution

The paper presents a novel boosting technique, DBNN, for classification with incomplete data, applicable to diverse problems including astronomical transient and spectral classification.

Findings

Applied to CRTS data with promising preliminary results

Successfully identified spectra using limited features

Demonstrated general applicability of the technique

Abstract

There are many occasions when one does not have complete information in order to classify objects into different classes, and yet it is important to do the best one can since other decisions depend on that. In astronomy, especially time-domain astronomy, this situation is common when a transient is detected and one wishes to determine what it is in order to decide if one must follow it. We propose to use the Difference Boosting Neural Network (DBNN) which can boost differences between feature vectors of different objects in order to differentiate between them. We apply it to the publicly available data of the Catalina Real-Time Transient Survey (CRTS) and present preliminary results. We also describe another use with a stellar spectral library to identify spectra based on a few features. The technique itself is more general and can be applied to a varied class of problems.