"Real or Junk?" will be back soon
08 August 2022, by Adam McMaster in Variable Stars
In November I wrote that we had added a large batch of 400,000 subjects to the Real or junk? workflow on SuperWASP Variable Stars. I'm pleased to report that all of those subjects have now been completed!
It's been about a year and a half since we launched the Real or junk? workflow, and in that time we've managed to filter over 1.1 million light curves, narrowing down the set to about 400,000 real subjects which don't look like random noise. This is amazing progress! This smaller set of subjects then feeds into the main light curve classification workflow, speeding up our search for variable stars.
We're not done yet, though. There is still plenty of data left to filter. But, we now have enough light curves which have been labelled as either real or junk that we can train a computer to identify them automatically. SuperWASP Variable Stars team member Hugh Dickinson has already built a computer model which can do that, and we've been testing it out in recent weeks.
The computer model is pretty good at classifying light curves and most of the time it can correctly separate real data from junk. However, since the data is pretty noisy we can't rely on this alone, as it does sometimes make mistakes and we don't want to miss anything.
That brings us to what's next for Real or junk? – we're going to put it on pause for a few weeks, and when we bring it back it will be powered by Hugh's computer model. We'll classify everything with the computer model first, and anything it identifies as real will be sent to the main workflow to be classified in full. For anything it identifies as junk, we'll ask you to give us one manual classification (rather than the three we ask for at the moment) and if you think it's real (i.e. if you disagree with the computer) we'll send it to be classified in the main workflow.
This way we can maintain the high level of accuracy we've had until now, but without needing as much work from our volunteers. We wouldn't want to waste your time by asking you to do something a computer can do. This also means we will be able to filter out the rest of the junk much more quickly, helping us to find interesting new variable stars even faster!
The SuperWASP project is currently funded and operated by Warwick University and Keele University, and was originally set up by Queen’s University Belfast, the Universities of Keele, St. Andrews and Leicester, the Open University, the Isaac Newton Group, the Instituto de Astrofisica de Canarias, the South African Astronomical Observatory and by STFC.
The Zooniverse project on SuperWASP Variable Stars is led by Andrew Norton (The Open University) and builds on work he has done with his former postgraduate students Les Thomas, Stan Payne, Marcus Lohr, Paul Greer, and Heidi Thiemann, and current postgraduate student Adam McMaster.
The Zooniverse project on SuperWASP Variable Stars was developed with the help of the ASTERICS Horizon2020 project. ASTERICS is supported by the European Commission Framework Programme Horizon 2020 Research and Innovation action under grant agreement n.653477
VeSPA was designed and developed by Adam McMaster as part of his postgraduate work. This work is funded by STFC, DISCnet, and the Open University Space SRA. Server infrastructure was funded by the Open University Space SRA.