Although ANN has become a modern-day buzzword, its roots go back more than 60 years. Stanford researchers created the first neural network to solve a real-world problem in the late 1950s to eliminate phone line echoes. However, it would be another three decades before the tool would be applied to astronomical research. 

AI algorithms highly support the attempt of man to understand the universe. For example, in 2015, a French research team used Hadoop, Hive, and Google's MapReduce for Large Synoptic Survey Telescope (LSST) data and query management. The following year, a team of US researchers built an Apache Spark-based toolkit called KIRA, which was deployed on Amazon's EC2 cloud.  

ISRO is not behind in using AI to study space and planets. ISRO's research on AI for observational purposes, satellite navigation, meteorology and space assistance started in November 2018. They published a synopsis to the universities regarding their study. Mentioned following are some of the ISRO initiatives that revolve around AI and ML.

A group of Caltech astronomers have recently used an ML algorithm to classify 1000 supernovae completely autonomously. The algorithm was applied to data captured by the Zwicky Transient facility or ZTF, a sky survey instrument based at Caltech's Palomar Observatory.  

What are supernovae? 

According to NASA, a supernova is the biggest explosion humans have ever seen. Each blast is the extremely bright, super-powerful explosion of a star. The "Last hurrah" of a dying massive star is a cause of one type of supernova. This occurs when a star at least five massive times the mass of the sun goes out with a massive bang. 

The second type of supernova can happen in systems where two stars orbit one another, and at least one of those stars is an earth-sized white dwarf. A white dwarf is what's left after a star the size of our sun has run out of fuel. If two white dwarfs collide with one another or a dwarf pulls too much matter from a nearby star, it can explode. 

Supernovae are very bright. They can outshine entire galaxies for a few days or even months and be seen across the universe. However, this phenomenon is rare. In fact, astronomers believe that about two or three supernovas occur each century in galaxies like our Milky Way.

Caltech's ML algorithm 

ZFT scans the night skies every night to look for changes called transient events. This change includes everything from moving asteroids to black holes. ZFT send out hundreds of alters a night to astronomers around the world. Astronomers then use a telescope to investigate the nature of the changing objects. The ZFT data has led to the discovery of thousands of supernovae so far. 

However, due to the massive load of data generated by ZFT, team members cannot sort through independently. Instead, they developed an ML algorithm to aid the searchers. They developed SNlascore for the task of classifying candidate supernovae. Currently, SNlascore can classify Type las supernovae which occur due to the dying of a star. In addition, it allows astronomers to measure the expansion of the universe.  

According to Christoffer Fremling, a staff astronomer at Caltech and the mastermind behind SNlascore, the algorithm is remarkably accurate. Furthermore, the Caltech astronomers state that this work demonstrates well how ML applications are coming of age in near real-time astronomy.