Humans are occasionally irreplaceable when it comes to finding new heavenly bodies because of their aptitude for pattern recognition. However, there are other situations where computers may detect things that people cannot see, such as a recent case where an exoplanet was found using machine learning.
Within a protoplanetary disc known as HD 142666, the exoplanet was found by University of Georgia astronomers. A protoplanetary disc is a spinning disc of gas that orbits newly-formed stars and is where planets are created. As matter gathers in these discs, planets are created as the gravitational attraction of the clumped material eventually draws in additional material. In order to explore for exoplanets that might have been overlooked the first time around, the researchers utilized a machine learning model to analyze a previous batch of observations of a large collection of protoplanetary discs. By observing the peculiar manner that gas travelled within a certain disc, they were able to pinpoint one disc where a planet was most likely to be.
They used conventional methods to confirm the planet, but their models guided them to conduct the simulations and revealed the precise location of the planet, according to lead author Jason Terry. They discovered a disc that had already been studied but didn’t have a planet when they applied their models to a collection of earlier data. Similar to other discoveries, scientists performed disc simulations and discovered that a planet might recreate the discovery.
The researchers claim that this is a proof of concept demonstrating how machine learning may be used to find new exoplanets even using data that has already been thoroughly examined. This might lead to the discovery of additional exoplanets in the future and to faster discovery rates.
This shows that their models, and machine learning in general, are capable of swiftly and precisely identifying crucial information that humans may overlook. According to Terry, this might significantly speed up processing and subsequent theoretical concepts. They believe there will be a significant role for these kinds of procedures as their data sets get even larger because it only took them about an hour to analyze the full catalogue and discover solid evidence for a new planet in a specific location.