Many of our imagined sci-fi futures pit humans and machines against each other – but what if they collaborated instead? Perhaps this is the future of astronomy.
As data sets get bigger and bigger, it becomes more and more difficult for small teams of researchers to analyze them. Scientists often turn to sophisticated machine learning algorithms, but they are not yet a substitute for human intuition and our brain’s superior pattern recognition abilities. However, a combination of the two can be the perfect team. Astronomers recently tested a machine learning algorithm that used information from volunteer citizen scientists to identify exoplanets in data from NASA’s Transiting Exoplanet Survey Satellite (TESS).
“This work shows the benefits of using machine learning with human participation,” Shresht Malik, a physicist at the University of Oxford in the UK and lead author of the publication, told Space.com.
Related: 10 Amazing Exoplanet Discoveries
The researchers used a typical machine learning algorithm known as a convolutional neural network. This computer algorithm looks at images or other information that people have correctly labeled (called “training data”) and learns to identify important features. Once trained, the algorithm can identify these features in new data that it has not seen before.
However, in order for the algorithm to work accurately, it needs a lot of this labeled training data. “It’s hard to get tags on this scale without the help of citizen scientists,” Nora Eisner, an astronomer at the Flatiron Institute in New York and co-author of the study, told Space.com.
People from all over the world have contributed by searching for and marking exoplanet transits through the Planet Hunters TESS project on Zooniverse, an online science research platform. Citizen science has the added benefit of “sharing the euphoria of discovery with non-scientists, promoting scientific literacy and public confidence in scientific research,” John Zink, an astronomer at the California Institute of Technology who is not affiliated with Space.com, told Space.com. this new study.
Finding exoplanets is a tough job – they’re tiny and dim compared to the massive stars they orbit. In data from telescopes such as TESS, astronomers can detect faint dips in a star’s light as a planet passes between it and the observatory, known as the transit method.
However, satellites oscillate in space, and stars are not perfect light bulbs, which sometimes makes it difficult to detect transits. Zink believes that partnering with machine learning “could greatly improve our ability to detect exoplanets” in such real-life, noisy data.
Some planets are harder to find than others. Planets with a long period of revolution around their star are rarer, which means a longer period of time between dips in the light. TESS only studies each section of the sky for a month, so only one transit can be recorded for these planets instead of several periodic changes.
“With citizen science, we are especially good at identifying long-lived planets, planets that tend to be overlooked in automated transit searches,” Eisner said.
The work has the potential to go well beyond exoplanets, Malik said, as machine learning is rapidly becoming a popular technique in many aspects of astronomy. “I see its impact only increasing as our datasets and methods get better.”
The research was presented at the Machine Learning and Physical Sciences Workshop at the 36th Conference on Neural Information Processing Systems (NeurIPS) in December and described in a paper hosted on the arXiv.org preprint server.
Follow the author at @briles_34 on Twitter. Follow us on Twitter @Spacedotcom and on Facebook.