The Kepler spacecraft, which operated from 2009 through 2018, observed 530,506 stars, “staring” at star fields to detect exoplanets which transit (pass in front of, as seen from Earth) the stars they orbit. So far, 2,662 transiting planets have been identified from the data it collected. The massive data set Kepler collected is still being mined by scientists and amateur planet hunters, who have continued to identify new planet candidates missed by the original analysis process.
Now, a paper, “ExoMiner: A Highly Accurate and Explainable Deep Learning Classifier that Validates 301 New Exoplanets”, reports that a deep learning classifier trained on the criteria used by astronomers to find transit candidates in the raw data set demonstrates 93.6% accuracy on a test data set (compared to 76.3% for the previously best automated classifier), and has discovered 301 new exoplanets in the Kepler data set. Here is the abstract.
The Kepler and TESS missions have generated over 100,000 potential transit signals that must be processed in order to create a catalog of planet candidates. During the last few years, there has been a growing interest in using machine learning to analyze these data in search of new exoplanets. Different from the existing machine learning works, ExoMiner, the proposed deep learning classifier in this work, mimics how domain experts examine diagnostic tests to vet a transit signal. ExoMiner is a highly accurate, explainable, and robust classifier that 1) allows us to validate 301 new exoplanets from the MAST Kepler Archive and 2) is general enough to be applied across missions such as the on-going TESS mission. We perform an extensive experimental study to verify that ExoMiner is more reliable and accurate than the existing transit signal classifiers in terms of different classification and ranking metrics. For example, for a fixed precision value of 99%, ExoMiner retrieves 93.6% of all exoplanets in the test set (i.e., recall=0.936) while this rate is 76.3% for the best existing classifier. Furthermore, the modular design of ExoMiner favors its explainability. We introduce a simple explainability framework that provides experts with feedback on why ExoMiner classifies a transit signal into a specific class label (e.g., planet candidate or not planet candidate).
The developers note that “explainability” was an important criterion in the design. This means that when the classifier detects a candidate, does not just go “bing”, but is rather able to explain why this candidate was selected, just as a human examining the data would do.