FEATURES
Open Science Cloud
▲ FIG . 1 :
A facial recognition algorithm would tell you where the faces are , but at the moment only a human has the ability to “ un-ask ” the question and say , wait a minute , there ’ s a clown in the image ! towards the Galactic centre as a signature of dark matter particle annihilation . This would be the natural location for such a signal , given the expected density peak in the dark matter distribution . However , the instrument team themselves responded to this claim ( Ackermann et al . 2017 , Astrophysical Journal 840-1 , article id . 43 ) by re-interpreting the claimed signal as an observational systematic . Without taking any view on the merits or otherwise of either side of this debate , it ' s clear that the usefulness of FAIR data will be limited by the supporting supplementary contextual information , beyond metadata even into training . The further that the data are from a user ' s subject specialism , the more curated the interaction must be with that data . The most extreme example of this is citizen science . In astronomy , there aren ’ t many serious consequences for misunderstanding and misuse of data , but in e . g . healthcare or climate science the stakes are much higher and the specialist communities have to take much more care .
But in astronomy at least , we are safer to experiment with bringing the public into open science . And so , my Open University colleagues Hugh Dickinson and James Pearson have been building Jupyter notebooks that demonstrate how to design , build and run citizen science projects , as well as fold in machine learning such as the excellent deep learning tools by Mike Walmsley . The goal is to have plug-and-play exemplars to help the professional community engage with both the EOSC open science tools and with the abundant research effort available from the volunteer community .
Our vision for a professional-amateur collaboration would work like this , at least in the context of the EOSC products and services for astronomy and astroparticle physics built by the EOSC ESCAPE project ( European Science Cluster of Astronomy & Particle Physics ESFRI Research Infrastructures , https :// projectescape . eu ). We start with the professional scientists working already on the ESCAPE science analysis platform , using data from its data lake . On that platform they are building and running data mining projects , initially deploying volunteers in citizen science . In many cases in astronomy , citizen science volunteers are already able to link out of the citizen science projects into the many accessible professional EOSC virtual observatory tools , armed with the new knowledge that they have acquired as part of the citizen science project that gives them the context of the scientific data . The professionals collate and ( where necessary ) reduce or clean the volunteer data on the science analysis platform , then use it as a training set for machine learning to accelerate the classifications . One can then use the machines to classify the most straightforward and unambiguous items to classify , and refocus the human effort on the difficult edge cases that are the most sensible use of human
22 EPN 54 / 2