A news and commentary website that is independent of academics and journalists. A news and commentary website that is independent of academics and journalists.Astronomy is all data. As the universe grows, so does our knowledge. The biggest challenge for the next generation in astronomy is how we are going to analyze all of the data.Astronomers have turned to machine learning and AI to tackle these problems. They are now able to quickly build new tools to search for the next great breakthroughs. Here are four ways AI can help astronomers.1. 1.There are many ways to find a planet. However, the best way is to study transits. An exoplanet that passes directly in front of its parent star blocks some of what we can see.Astronomers can build a picture of the light dips by observing multiple orbits of an exoplanet. This allows them to identify planets such as their mass, size, and distance from their star. This technique was used by the Nasas Kepler space telescope to great success. They viewed thousands of stars simultaneously, looking out for signs that planets are causing dips.We see the light drop when planets pass in front of their stars. Image via NASA/Tim PyleAlthough humans are quite skilled at spotting these dips, it is a skill that takes time. Humans are unable to keep up with the increasing number of missions dedicated to finding new exoplanets like Nasas (Transiting Exoplanets Survey Satellite). AI is here to help.Combining time-series analysis techniques that analyse data in a sequential order with time has been combined with an AI type to identify signals from exoplanets up to 96% accurate.2. 2. Gravitational wavesTime-series models can be used to find exoplanets. They are also great for finding signals from the most devastating events in the universe, such as mergers between neutron stars and black holes.These extremely dense bodies can fall inwards and emit ripples in space-time. This can be detected here by measuring faint signals. Machine learning has helped identify the signals from dozens of these events. Ligo and Virgo are both collaborators in gravity wave detector collaborations.The Ligo and Virgo teams can train models using simulated data from black hole mergers. This allows them to identify potential events in a matter of seconds and then send alerts to astronomers all over the globe to direct their telescopes in the correct direction.3. The changing skyThe Vera Rubin Observatory is currently in Chile and will begin to collect over 80 Terabytes of images every night. This will allow you to see the changes in stars and galaxies as time passes. One terabyte equals 8,000,000,000,000 bits.Rubin's Legacy Survey of Space and Time will record and process hundreds of petabytes of information over the course of its planned operations. For context, 100 petabytes refers to the amount of space required to store each photo on Facebook or 700 years worth of high-definition video.It won't be possible to log on to the servers and access that data. Even if you could, it wouldn't be possible to locate what you are looking for.These next-generation surveys will be searched using machine learning techniques to highlight important data. One algorithm may search for rare events like supernovae, which are dramatic explosions at end of stars lives. Another algorithm might search for quasars. The team can train computers to recognize signals from particular astronomical phenomena and get the correct data to the right people.4. 4. Gravitational lensesWe collect ever more data about the universe and sometimes have to discard data that isn't necessary. How can we discover the most rare objects among these vast amounts of data?Strong gravitational lenses are a celestial phenomenon that excites many astronomers. This happens when two galaxies align along our line-of-sight. The closest galaxy's gravity acts like a lens, magnifying the distant object and creating rings, crosses, and double images.The blue ring is a light source from a distant galaxy that has been distorted by the central red galaxy. Image via NASA/ESA/Hubble, CC BYThese lenses are like looking for a needle in the middle of a haystack. This haystack is the size of the entire observable universe. As we acquire more images of galaxies, it will only get more difficult.Astronomers around the globe participated in the Strong Gravitational Lens Finding Challenge 2018. They competed to find the best algorithm to automatically locate these lenses.This challenge was won by a convolutional neural net, which can learn to deconstruct images using different filters so it can classify them as lens-containing or not. Surprisingly these models outperformed people by detecting subtle differences in images that humans can't see.Astronomers will be able to collect petabytes (thousands of terabytes) of data over the next decade using new instruments such as the Vera Rubin Observatory. Astronomers will increasingly rely upon machine-learning techniques as they look deeper into the universe.This article is by Ashley Spindler (Research Fellow, Astrophysics University of Hertfordshire), and has been republished from The Conversation using a Creative Commons License. You can read the original article.
