Machine learning algorithm discovers distant quasar

Machine Learning


Artificial intelligence helps scientists find potential quasars from the early universe by using specially trained machine learning algorithms to process telescope images and look for images distorted by gravitational lensing.

Machine learning algorithms have helped scientists search for quasars. Credit: phys.org

Quasars and gravitational lensing

The first indication that an object is a quasar is its red color, which is confirmed by observing its individual spectrum, but some candidates at high redshifts have their appearance distorted by gravitational lensing and are therefore erroneously excluded from further study.

This phenomenon occurs when a massive object, such as a galaxy, sits between us and a distant object: the galaxy's mass bends space, acting like a magnifying glass, bending the path of light from the distant object and creating a distorted image.

This arrangement can be useful, because gravitational lensing magnifies the image of the quasar, making it brighter and easier to detect. But it can also deceptively change the quasar's appearance: interference from light from stars in the intervening lensing galaxy can make the quasar appear bluer, and the curvature of space-time can make the quasar appear fuzzy or faded. Both of these effects make it a likely quasar candidate.

So a team of astronomers led by Xander Byrne, an astronomer at the University of Cambridge and lead author of the paper that published the results in the journal Nature, Monthly Bulletin of the Royal Astronomical Societydecided to recover a lensed quasar that had been overlooked in previous studies.

Searching for quasars distorted by gravitational lensing

Byrne is the author of the vast Dark Energy Survey (death) Data archive. death The research was conducted using the Dark Energy Camera on the National Science Foundation's Cerro Tololo Inter-American Observatory's 4-meter Victor M. Blanco Telescope as part of the NSF NOIRLab program.

The challenge was to find these cosmic gems in the vast ocean of data.

Complete death The dataset contains more than 700 million objects. Byrne pared down this archive by comparing the data with images from other surveys to eliminate unlikely candidates, such as objects that might be brown dwarfs. Brown dwarfs are completely different in almost every way from quasars, but can look strikingly similar to quasars in images. This process resulted in a much more manageable dataset containing 7,438 objects.

Byrne needed to be as efficient as possible in searching for these 7,438 objects, but he knew that using traditional methods would likely miss the high-redshift lensed quasars he was looking for. To avoid prematurely ruling out lensed quasars, the scientists used a contrast learning algorithm.

Machine learning algorithm searches for quasars

Contrast learning is a type of artificial intelligence (AI) algorithm that uses sequential decisions to assign each data point to a group depending on what it is or isn't. Byrne's decision not to rely on human visual interpretation led to the idea of ​​an unsupervised AI process, where the algorithm itself, rather than a human, drives the learning process.

Supervised machine learning algorithms are based on so-called ground truth defined by a human programmer: for example, the process starts with a description of a cat and then proceeds to decide “this is/is not a photo of a cat, this is/is not a photo of a black cat”.

In contrast, unsupervised algorithms don't rely on a human-given definition as the basis for their decisions. Instead, the algorithm sorts each data point by its similarity to other data points in the set. In this case, the machine might find similarities between images of multiple animals and group them into cats, dogs, giraffes, penguins, etc.

Starting with 7,438 Burn objects, the algorithm classified the objects into groups in an unsupervised fashion. Using a geographic analogy, the team called the data groups archipelagos. Within them, small “island” subsets of objects were grouped together as potential quasars. Among these candidates, four stood out like pearls in a pile of stones.

Using archived data from the Gemini South Telescope at the Gemini International Observatory, NSF NOIR LabByrne confirmed that three of the four candidates in “Quasar Island” are indeed high-redshift quasars, and one of them is very likely to be a gravitationally lensed high-redshift quasar, the cosmic discovery Byrne was looking for. The team is now planning further imaging to confirm the nature of the quasar.

Based on phys.org material





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *