TY - JOUR
T1 - Galaxy Zoo
T2 - Probabilistic morphology through Bayesian CNNs and active learning
AU - Walmsley, Mike
AU - Smith, Lewis
AU - Lintott, Chris
AU - Gal, Yarin
AU - Bamford, Steven
AU - Dickinson, Hugh
AU - Fortson, Lucy
AU - Kruk, Sandor
AU - Masters, Karen
AU - Scarlata, Claudia
AU - Simmons, Brooke
AU - Smethurst, Rebecca
AU - Wright, Darryl
N1 - Publisher Copyright:
© 2019 The Author(s).
PY - 2020/1/1
Y1 - 2020/1/1
N2 - We use Bayesian convolutional neural networks and a novel generative model of Galaxy Zoo volunteer responses to infer posteriors for the visual morphology of galaxies. Bayesian CNN can learn from galaxy images with uncertain labels and then, for previously unlabelled galaxies, predict the probability of each possible label. Our posteriors are well-calibrated (e.g. for predicting bars, we achieve coverage errors of 11.8 per cent within a vote fraction deviation of 0.2) and hence are reliable for practical use. Further, using our posteriors, we apply the active learning strategy BALD to request volunteer responses for the subset of galaxies which, if labelled, would be most informative for training our network. We show that training our Bayesian CNNs using active learning requires up to 35-60 per cent fewer labelled galaxies, depending on the morphological feature being classified. By combining human and machine intelligence, Galaxy zoo will be able to classify surveys of any conceivable scale on a time-scale of weeks, providing massive and detailed morphology catalogues to support research into galaxy evolution.
AB - We use Bayesian convolutional neural networks and a novel generative model of Galaxy Zoo volunteer responses to infer posteriors for the visual morphology of galaxies. Bayesian CNN can learn from galaxy images with uncertain labels and then, for previously unlabelled galaxies, predict the probability of each possible label. Our posteriors are well-calibrated (e.g. for predicting bars, we achieve coverage errors of 11.8 per cent within a vote fraction deviation of 0.2) and hence are reliable for practical use. Further, using our posteriors, we apply the active learning strategy BALD to request volunteer responses for the subset of galaxies which, if labelled, would be most informative for training our network. We show that training our Bayesian CNNs using active learning requires up to 35-60 per cent fewer labelled galaxies, depending on the morphological feature being classified. By combining human and machine intelligence, Galaxy zoo will be able to classify surveys of any conceivable scale on a time-scale of weeks, providing massive and detailed morphology catalogues to support research into galaxy evolution.
KW - Galaxies: evolution
KW - Galaxies: statistics
KW - Galaxies: structure
KW - Methods: data analysis
KW - Methods: statistical
UR - http://www.scopus.com/inward/record.url?scp=85079671660&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85079671660&partnerID=8YFLogxK
U2 - 10.1093/mnras/stz2816
DO - 10.1093/mnras/stz2816
M3 - Article
AN - SCOPUS:85079671660
SN - 0035-8711
VL - 491
SP - 1554
EP - 1574
JO - Monthly Notices of the Royal Astronomical Society
JF - Monthly Notices of the Royal Astronomical Society
IS - 2
ER -