Stellar Classification vis-à-vis Convolutional Neural Network

Abstract

As a result of recent advancements in technology, a variety of new computational fields have emerged. Some examples of these fields are machine learning and intelligence, information science, the internet of things, and others. The advancement of humanity will be greatly aided by these fields. The development of Artificial Intelligence led to the creation of a great deal of Neural Networks. Convolutional Neural Networks are one variation of Neural Networks that we are utilizing in this work. These networks are known to perform quite admirably for Image Categorization, which is one of the purposes for which we are utilizing them. The work encompasses Stellar Classification. There are many stellar entities occupying the region known as universal space. Astrophysicists are familiar with a good number of them, but there are still a great many of these types of entities that have not been discovered yet. Because of the great distance that separates our planet from other stellar entities, any attempt to communicate with them through any channel is highly unlikely to be successful. The most information we could possibly acquire is just a guess as to what kind of entity they are. So, if any scientific observatory comes with a nascent search of any distant entity, we could potentially predict which stellar group they belong to. For the purposes of this work, we are only going to focus on two different types of Stella: Stars and Galaxies. For the purpose of training the Convolutional Neural Network, we have used a dataset on Stellar Types with Image Categorization created by the Aryabhata Research Institute of Observational Sciences (ARIES), which is located in Nainital, India. This dataset was made publicly available.