APPLICATION OF DEEP LEARNING NEURAL NETWORK FOR CLASSIFICATION OF BIG DATA OF ASTRONOMIC OBSERVATIONS

Purpose: In the process of astronomical observations vast amounts of data are collected. The BSA (Big Scanning Antenna) used in the study of impulse phenomena, daily logs 87.5 GB of data (32 TB per year). The aim of this work is to develop the web-service which assists the experts with classification of new astronomic observations. The Azure Machine Learning Studio which offers a Deep Neural Network algorithm is used as a tool for web-service developing. Design/methodology/approach: Experts classified 83096 individual observations (on the segment of the study July 2012 – October 2013). Over 89 % of the sample correspond to pulsars, twinkling springs and rapid radiotransmitters, and all other classes of observations belong to hardware failures, interference, the flight of the Earth satellite and aircraft. There were allocated 15 classes of observations. Findings: Such a sample, divided into classes allows using the machine learning algorithms. It has become possible to develop an automated service for short-term/long-term monitoring of various classes of radio sources (including radiotransmitted of different nature), monitoring the Earth’s ionosphere, the interplanetary and the interstellar plasma, search and monitoring of different classes of radio sources. Monitoring in this case refers to the automatic filtering and detection of earlier unclassified impulse phenomena. Currently, for automatic filtering, statistical analysis methods are used. This paper considers an alternative method supposed to be using neural network machine learning algorithm which processes the input into raw data, and after processing by the hidden layer through the output layer determines the class of pulse phenomena. Сonclusions: Creating a neural network model, trained on a sample and classifying earlier unclassified impulse phenomena, is performed using the cloud service Microsoft Azure Machine Learning Studio. The Web service, created based on the aforesaid model, allows classifying both single impulse phenomena in real time (Request / Reply) and data sampling for a certain period (Batch processing).

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.