During the National Astronomy Meeting held this week at The University of Warwick in the UK, graduate student Eleonora Parrag presented her results on using machine learning to build artificial supernovae spectra.
Supernovae evolve rapidly within a few hundred days of their parent star’s explosion. If we’re lucky, we can spot these supernovae early on in the astrophotographs of community astronomers and photographers. Then we can collect data in the days following to see how the supernova evolves. Collecting spectra will help determine the chemical composition of the supernova and possibly even reveal some of the conditions involved in the stellar explosion. But we don’t always have enough telescope time or data to get a full set of spectra.
Enter machine learning. By training an algorithm on the observations of well-documented supernovae, Parrag was able to reproduce the features for less documented supernovae, essentially filling in the gaps in the spectra.
The spectra can then be used to understand the physics of supernovae and possibly find patterns within their populations. Parrag explains: Machine learning can help us find patterns and potentially even new ideas in physics in the huge amounts of data from supernovae we can observe now and in the foreseeable future.
Sometimes we use machine learning when we have too much data, and sometimes, we use it when we don’t have enough. Either way, pretty amazing stuff.
RAS press release