DETECTING BARS IN GALAXIES USING CONVOLUTIONAL NEURAL NETWORKS

Recently astronomers made a fascinating discovery about the existence of black hole near the center of our solar system. Discoveries like these will be more frequent in future as the astronomers are able to gain huge amounts of data everyday. Neural networks which are known to handle huge amount of data will have a great relevance in this context. We will demonstrate an application on how to detect bar like structure in galaxies. Convolutional Neural Network (CNN) is a popular deep learning technique inspired by the natural visual perception mechanism of the living organisms. It can be used to handle huge volumes of data.This article makes a survey of several steps we need to go through in order to build a convolutional neural network. The various steps involved are convolution operation, Relu layer, max pooling, Flattening step and finally the full connection. These steps will be applied to detect whether the galaxy is barred or unbarred. We will be using a sample of 9346 galaxies in the redshift range from 0.009-0.2 from sloan digital sky survey having 3864 barred galaxies and the rest being unbarred. We were able to obtain a top accuracy of 95.68 in identifying the bars in galaxy images using the trained network. This survey will help all understand on how those final neurons understand how to classify image. We will also review the efforts to understand CNNs and review major applications of CNNs in computer vision tasks.