Radio Galaxy Zoo: Building a multi-purpose foundation model for radio astronomy with self-supervised learning. (arXiv:2305.16127v1 [astro-ph.IM])

Radio Galaxy Zoo: Building a multi-purpose foundation model for radio astronomy with self-supervised learning. (arXiv:2305.16127v1 [astro-ph.IM])
<a href="http://arxiv.org/find/astro-ph/1/au:+Slijepcevic_I/0/1/0/all/0/1">Inigo V. Slijepcevic</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Scaife_A/0/1/0/all/0/1">Anna M. M. Scaife</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Walmsley_M/0/1/0/all/0/1">Mike Walmsley</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Bowles_M/0/1/0/all/0/1">Micah Bowles</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Wong_O/0/1/0/all/0/1">O. Ivy Wong</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+Shabala_S/0/1/0/all/0/1">Stanislav S. Shabala</a>, <a href="http://arxiv.org/find/astro-ph/1/au:+White_S/0/1/0/all/0/1">Sarah V. White</a>

In this work, we apply self-supervised learning with instance differentiation
to learn a robust, multi-purpose representation for use in radio astronomy. We
exceed baseline supervised classification performance by a statistically
significant margin for most label volumes in the in-distribution classification
case and for all label volumes in the out-of-distribution case, with our model
reducing the test set error by up to 5% depending on label volume: from 5% to
2%. Our model is also able to maintain high classification accuracy with very
few labels, with only 7.79% error when only using 145 labels. We further
demonstrate that by using our foundation model, users can efficiently trade off
compute, human labelling cost and test set accuracy according to their
respective budgets, allowing for efficient classification in a wide variety of
scenarios. Visualizations of our labelled and un-labelled data show that our
model’s representation space is structured with respect to physical properties
of the sources, such as angular source extent. We show that the learned
representation is scientifically useful even if no labels are available by
performing a similarity search, finding hybrid sources in the RGZ DR1 data-set
without any labels. We show that good augmentation design and hyper-parameter
choice can help achieve peak performance, while emphasising that optimal
hyper-parameters are not required to obtain benefits from self-supervised
pre-training.

In this work, we apply self-supervised learning with instance differentiation
to learn a robust, multi-purpose representation for use in radio astronomy. We
exceed baseline supervised classification performance by a statistically
significant margin for most label volumes in the in-distribution classification
case and for all label volumes in the out-of-distribution case, with our model
reducing the test set error by up to 5% depending on label volume: from 5% to
2%. Our model is also able to maintain high classification accuracy with very
few labels, with only 7.79% error when only using 145 labels. We further
demonstrate that by using our foundation model, users can efficiently trade off
compute, human labelling cost and test set accuracy according to their
respective budgets, allowing for efficient classification in a wide variety of
scenarios. Visualizations of our labelled and un-labelled data show that our
model’s representation space is structured with respect to physical properties
of the sources, such as angular source extent. We show that the learned
representation is scientifically useful even if no labels are available by
performing a similarity search, finding hybrid sources in the RGZ DR1 data-set
without any labels. We show that good augmentation design and hyper-parameter
choice can help achieve peak performance, while emphasising that optimal
hyper-parameters are not required to obtain benefits from self-supervised
pre-training.

http://arxiv.org/icons/sfx.gif