A Dynamic, Modular Intelligent-Agent Framework for Astronomical Light Curve Analysis and Classification

Modern time-domain astronomy is capable of collecting a staggeringly large amount of data on millions of objects in real time. This makes it almost impossible for objects to be identified manually. Therefore the production of methods and systems for the a

  • PDF / 1,042,811 Bytes
  • 12 Pages / 439.37 x 666.14 pts Page_size
  • 89 Downloads / 198 Views

DOWNLOAD

REPORT


1

Applied Computing Research Group, Faculty of Engineering and Technology, Liverpool John Moores University, Byrom Street, Liverpool L3 3AF, UK [email protected], [email protected], {D.Aljumeily,A.Hussain,P.Fergus}@ljmu.ac.uk 2 Astrophysics Research Institute, IC2, Liverpool Science Park, Liverpool John Moores University, 146 Brownlow Hill, Liverpool L3 5RF, UK [email protected]

Abstract. Modern time-domain astronomy is capable of collecting a stagger‐ ingly large amount of data on millions of objects in real time. This makes it almost impossible for objects to be identified manually. Therefore the production of methods and systems for the automated classification of time-domain astronom‐ ical objects is of great importance. The Liverpool Telescope has a number of wide-field image gathering instruments mounted upon its structure. Utilizing a database established by a pre-processing operation upon these images, containing millions of candidate variable stars with multiple time-varying magnitude obser‐ vations, we applied a method designed to extract time-translation invariant features from the time-series light curves. These efforts were met with limited success due to noise and uneven sampling within the time-series data. Addition‐ ally, finely surveying these light curves is a processing intensive task. Fortunately, these algorithms are capable of multi-threaded implementations based on avail‐ able resources. Therefore we propose a new system designed to utilize multiple intelligent agents that distribute the data analysis across multiple machines whilst simultaneously a powerful intelligence service operates to constrain the light curves and eliminate false signals due to noise and local alias periods. Keywords: Data analysis methods · Big data mining · Machine learning · Uneven time-series analysis · Light curve analysis · Variable stars · Binary stars · Harmonic regression · Harmonic feature extraction · Multi-agent systems · Period detection

1

Introduction

Astronomy is entering a period of unprecedented data gathering capability. Advances in observational, storage and data processing technologies have allowed for extended sky surveys such as the Sloan Digital Sky Survey (SDSS) to be conducted and exploited [1]. Within the next decade a number of even larger surveys are also planned. Tech‐ nology is now at a point where it has become possible to gather data on wide regions of the sky repeatedly over variable time periods [2]. This data can be analyzed for periodic © Springer International Publishing Switzerland 2016 D.-S. Huang et al. (Eds.): ICIC 2016, Part I, LNCS 9771, pp. 820–831, 2016. DOI: 10.1007/978-3-319-42291-6_81

A Dynamic, Modular Intelligent-Agent Framework

821

structure which can then be used to create physical models by fitting weighted regression learning algorithms. These methods can provide us with valuable knowledge about the presence and classification of astronomical objects that are periodically changing in time as well as identifying transient phenomena [3]. Time d