arXiv subjects

Anna Scaife

Publications and source records attributed to Anna Scaife.

At least 19 recordsLinked to original sources

arXiv2026

Deep Learning for Astrophysics: An Open Textbook from the NASA Cosmic Origins AI/ML Science and Technology Interest Group

Recent community assessments identify education as a principal barrier to adopting modern machine learning in astronomy. We present Deep Learning for Astrophysics, a freely available textbook at https://deeplearning4astro.com, curated from the NASA Cosmic Origins Artificial Intelligence and Machine Learning Science and Technology Interest Group (AI/ML STIG) lecture series. The book collects 23 chapters by 17 lecturers across six parts, moving from computational foundations and deep-learning architectures through generative modeling, simulation-based inference, reinforcement learning, and large-language-model agents to the practice of AI-laden science. Many include executable notebooks using astronomical data.

astro-ph.IM↗

arXiv2025

Intrinsic Dimension Estimation for Radio Galaxy Zoo using Diffusion Models

In this work, we estimate the intrinsic dimension (iD) of the Radio Galaxy Zoo (RGZ) dataset using a score-based diffusion model. We examine how the iD estimates vary as a function of Bayesian neural network (BNN) energy scores, which measure how similar the radio sources are to the MiraBest subset of the RGZ dataset. We find that out-of-distribution sources exhibit higher iD values, and that the overall iD for RGZ exceeds those typically reported for natural image datasets. Furthermore, we analyse how iD varies across Fanaroff-Riley (FR) morphological classes and as a function of the signal-to-noise ratio (SNR). While no relationship is found between FR I and FR II classes, a weak trend toward higher SNR at lower iD. Future work using the RGZ dataset could make use of the relationship between iD and energy scores to quantitatively study and improve the representations learned by various self-supervised learning algorithms.

cs.LG↗

arXiv2025

IRIS: A Bayesian Approach for Image Reconstruction in Radio Interferometry with expressive Score-Based priors

Inferring sky surface brightness distributions from noisy interferometric data in a principled statistical framework has been a key challenge in radio astronomy. In this work, we introduce Imaging for Radio Interferometry with Score-based models (IRIS). We use score-based models trained on optical images of galaxies as an expressive prior in combination with a Gaussian likelihood in the uv-space to infer images of protoplanetary disks from visibility data of the DSHARP survey conducted by ALMA. We demonstrate the advantages of this framework compared with traditional radio interferometry imaging algorithms, showing that it produces plausible posterior samples despite the use of a misspecified galaxy prior. Through coverage testing on simulations, we empirically evaluate the accuracy of this approach to generate calibrated posterior samples.

astro-ph.IM↗

arXiv2023

MCMC to address model misspecification in Deep Learning classification of Radio Galaxies

The radio astronomy community is adopting deep learning techniques to deal with the huge data volumes expected from the next-generation of radio observatories. Bayesian neural networks (BNNs) provide a principled way to model uncertainty in the predictions made by deep learning models and will play an important role in extracting well-calibrated uncertainty estimates from the outputs of these models. However, most commonly used approximate Bayesian inference techniques such as variational inference and MCMC-based algorithms experience a "cold posterior effect (CPE)", according to which the posterior must be down-weighted in order to get good predictive performance. The CPE has been linked to several factors such as data augmentation or dataset curation leading to a misspecified likelihood and prior misspecification. In this work we use MCMC sampling to show that a Gaussian parametric family is a poor variational approximation to the true posterior and gives rise to the CPE previously observed in morphological classification of radio galaxies using variational inference based BNNs.

astro-ph.IM↗

arXiv2023

Square Kilometre Array : Processing Voluminous MeerKAT Data on IRIS

Processing astronomical data often comes with huge challenges with regards to data management as well as data processing. MeerKAT telescope is one of the precursor telescopes of the World's largest observatory - Square Kilometre Array. So far, MeerKAT data was processed using the South African computing facility i.e. IDIA, and exploited to make ground-breaking discoveries. However, to process MeerKAT data on UK's IRIS computing facility requires new implementation of the MeerKAT pipeline. This paper focuses on how to transfer MeerKAT data from the South African site to UK's IRIS systems for processing. We discuss about our RapifXfer Data transfer framework for transferring the MeerKAT data from South Africa to the UK, and the MeerKAT job processing framework pertaining to the UK's IRIS resources.

astro-ph.IM↗

arXiv2022

Revision of Faraday rotation measure constraints on the primordial magnetic field using the IllustrisTNG simulation

Previously derived Faraday rotation constraints on the volume-filling intergalactic magnetic field (IGMF) have used analytic models that made a range of simplifying assumptions about magnetic field evolution in the intergalactic medium and did not consider the effect of baryonic feedback on large-scale structures. In this work we revise existing Faraday rotation constraints on the IGMF using a numerical model of the intergalactic medium from the IllustrisTNG cosmological simulation that includes a sophisticated model of the baryonic feedback. We use the IllustrisTNG model to calculate the rotation measure and compare the resulting mean and median of the absolute value of the rotation measure with data from the NRAO VLA Sky Survey (NVSS). The numerical model of the intergalactic medium includes a full magneto-hydrodynamic model of the compressed primordial magnetic field as well as a model of the regions where the magnetic field is not primordial, but is rather produced by the process of baryonic feedback. Separating these two types of regions, we are able to assess the influence of the primordial magnetic field on the Faraday rotation signal. We find that by correcting for regions of compressed primordial field and accounting for the fact that part of the intergalactic medium is occupied by magnetic fields spread by baryonic feedback processes rather than by the primordial field relaxes the Faraday rotation bound by a factor of $\simeq 3$. This results in $B_0<1.8\times10^{-9}$ G for large correlation length IGMFs.

astro-ph.CO↗

arXiv2022

A Compressed Sensing Faraday Depth Reconstruction Framework for the MeerKAT MIGHTEE-POL Survey

In this work we present a novel compute framework for reconstructing Faraday depth signals from noisy and incomplete spectro-polarimetric radio datasets. This framework is based on a compressed-sensing approach that addresses a number of outstanding issues in Faraday depth reconstruction in a systematic and scaleable manner. We apply this framework to early-release data from the MeerKAT MIGHTEE polarisation survey.

astro-ph.IM↗

arXiv2022

The contribution of magnetized galactic outflows to extragalactic Faraday rotation

Galactic outflows driven by star formation and active galactic nuclei blow bubbles into their local environments, causing galactic magnetic fields to be carried into intergalactic space. We explore the redshift-dependent effect of these magnetized bubbles on the Faraday Rotation Measure (RM) of extragalactic radio sources. Using the IllustrisTNG cosmological simulations, we separate the contribution from magnetic bubbles from that of the volume-filling magnetic component expected to be due to the seed field originating in the Early Universe. We use this separation to extract the redshift dependence of each component and to compare TNG model predictions with observation measurements of the NRAO VLA Sky Survey (NVSS). We find that magnetized bubbles provide a sizeable contribution to the extragalactic RM, with redshift-independent $\langle |{\rm RM}| \rangle \simeq 13$ rad/m$^2$ for sources at redshifts $z\ge 2$. This is close to the mean residual RM of $16$ rad/m$^2$ found from NVSS data in this redshift range. Using the IllustrisTNG simulations, we also evaluate a simple model for the contribution to residual RM from individual host galaxies and show that this contribution is negligible at high-redshift. While the contribution from magnetic bubbles in the IllustrisTNG model is currently compatible with observational measurements of residual RM, the next-generation RM sky surveys, which will be free from the wrapping uncertainty, have larger statistics and better sensitivity should be able to observe predicted flat contribution from magnetic bubbles at large redshifts. This should allow to experimentally probe magnetic bubbles and check models of galaxy feedback in cosmological simulations.

astro-ph.CO↗

arXiv2021

Quantum Machine Learning for Radio Astronomy

In this work we introduce a novel approach to the pulsar classification problem in time-domain radio astronomy using a Born machine, often referred to as a quantum neural network. Using a single-qubit architecture, we show that the pulsar classification problem maps well to the Bloch sphere and that comparable accuracies to more classical machine learning approaches are achievable. We introduce a novel single-qubit encoding for the pulsar data used in this work and show that this performs comparably to a multi-qubit QAOA encoding.

quant-ph↗

arXiv2021

E(2) Equivariant Self-Attention for Radio Astronomy

In this work we introduce group-equivariant self-attention models to address the problem of explainable radio galaxy classification in astronomy. We evaluate various orders of both cyclic and dihedral equivariance, and show that including equivariance as a prior both reduces the number of epochs required to fit the data and results in improved performance. We highlight the benefits of equivariance when using self-attention as an explainable model and illustrate how equivariant models statistically attend the same features in their classifications as human astronomers.

astro-ph.IM↗

arXiv2021

Weight Pruning and Uncertainty in Radio Galaxy Classification

In this work we use variational inference to quantify the degree of epistemic uncertainty in model predictions of radio galaxy classification and show that the level of model posterior variance for individual test samples is correlated with human uncertainty when labelling radio galaxies. We explore the model performance and uncertainty calibration for a variety of different weight priors and suggest that a sparse prior produces more well-calibrated uncertainty estimates. Using the posterior distributions for individual weights, we show that signal-to-noise ratio (SNR) ranking allows pruning of the fully-connected layers to the level of 30% without significant loss of performance, and that this pruning increases the predictive uncertainty in the model. Finally we show that, like other work in this field, we experience a cold posterior effect. We examine whether adapting the cost function in our model to accommodate model misspecification can compensate for this effect, but find that it does not make a significant difference. We also examine the effect of principled data augmentation and find that it improves upon the baseline but does not compensate for the observed effect fully. We interpret this as the cold posterior effect being due to the overly effective curation of our training sample leading to likelihood misspecification, and raise this as a potential issue for Bayesian deep learning approaches to radio galaxy classification in future.

astro-ph.IM↗

arXiv2020

Observations of magnetic fields surrounding LkH$α$ 101 taken by the BISTRO survey with JCMT-POL-2

We report the first high spatial resolution measurement of magnetic fields surrounding LkH$α$ 101, a part of the Auriga-California molecular cloud. The observations were taken with the POL-2 polarimeter on the James Clerk Maxwell Telescope within the framework of the B-fields In Star-forming Region Observations (BISTRO) survey. Observed polarization of thermal dust emission at 850 $μ$m is found to be mostly associated with the red-shifted gas component of the cloud. The magnetic field displays a relatively complex morphology. Two variants of the Davis-Chandrasekhar-Fermi method, unsharp masking and structure function, are used to calculate the strength of magnetic fields in the plane of the sky, yielding a similar result of $B_{\rm POS}\sim 115$ $\mathrmμ$G. The mass-to-magnetic-flux ratio in critical value units, $λ\sim0.3$, is the smallest among the values obtained for other regions surveyed by POL-2. This implies that the LkH$α$ 101 region is sub-critical and the magnetic field is strong enough to prevent gravitational collapse. The inferred $δB/B_0\sim 0.3$ implies that the large scale component of the magnetic field dominates the turbulent one. The variation of the polarization fraction with total emission intensity can be fitted by a power-law with an index of $α=0.82\pm0.03$, which lies in the range previously reported for molecular clouds. We find that the polarization fraction decreases rapidly with proximity to the only early B star (LkH$α$ 101) in the region. The magnetic field tangling and the joint effect of grain alignment and rotational disruption by radiative torques are potential of explaining such a decreasing trend.

astro-ph.GA↗

arXiv2019

Efficient Source Finding for Radio Interferometric Images

Object detection in astronomical images, generically referred to as source finding, is often performed before the object characterisation stage in astrophysical processing work flows. In radio astronomy, source finding has historically been performed by bespoke off-line systems; however, modern data acquisition systems as well as those proposed for upcoming observatories such as the Square Kilometre Array (SKA), will make this approach unfeasible. One area where a change of approach is particularly necessary is in the design of fast imaging systems for transient studies. This paper presents a number of advances in accelerating and automating the source finding in such systems.

astro-ph.IM↗

arXiv2018

Anomalous microwave emission from spinning nanodiamonds around stars

Several interstellar environments produce 'anomalous microwave emission', with brightness-peaks at tens-of-gigahertz frequencies. The emission's origins are uncertain - rapidly-spinning nano-particles could emit electric-dipole radiation, but polycyclic aromatic hydrocarbons proposed as the carrier are now found not to correlate with Galactic signals. The difficulty is to identify co-spatial sources over long lines of sight. Here we identify anomalous microwave emission in three proto-planetary discs. These are the only known systems that host hydrogenated nano-diamonds, in contrast to very common detection of polycyclic aromatic hydrocarbons. Spectroscopy locates the nano-diamonds close to the host-stars, at physically-constrained temperatures. Developing disc models, we reproduce the emission with diamonds 0.75-1.1 nanometres in radius, holding less than or equal to 1-2 per cent of the carbon budget. The microwave-emission:stellar-luminosity ratios are approximately constant, allowing nano-diamonds to be ubiquitous but emitting below detection thresholds in many star-systems. This can unify the findings with similar-sized diamonds found within solar system meteorites. As nano-diamond spectral absorption is seen in interstellar sightlines, these particles are also a candidate for generating galaxy-scale anomalous microwave emission.

astro-ph.SR↗

arXiv2018

The First VLBI Detection of a Spiral DRAGN Core

The existence of spiral DRAGNs challenges standard galaxy formation theories. We present the first observation of 0313$-$192, the archetypal spiral DRAGN, at VLBI resolutions. Spiral DRAGNs are Double Radio Sources Associated with Galactic Nuclei (DRAGNs) that are hosted by spiral galaxies. 0313$-$192 is an edge-on spiral galaxy that appears to host a 360 kpc double-lobed radio source. The core of this galaxy is clearly detected at L, S, and X-bands using the VLBA, signifying an ongoing active nucleus in the galaxy. This rules out the possibility that the spiral DRAGN is merely a chance alignment. The radio core has L$_{1.4\,GHz} \sim 3.0 \times 10^{23}$W Hz$^{-1}$. Radio components are detected to the South-West of the core, but there are no detections of a counterjet. Assuming a symmetric, relativistic jet, we estimate an upper limit to the inclination angle of $θ\lesssim 72$ degrees. The VLBI-detected radio jet components are extremely well-aligned with the larger-scale radio source suggesting little to no jet disruption or interaction with the ISM of the host galaxy.

astro-ph.GA↗

arXiv2015

Using SKA Rotation Measures to Reveal the Mysteries of the Magnetised Universe

We know that magnetic fields are pervasive across all scales in the Universe and over all of cosmic time and yet our understanding of many of the properties of magnetic fields is still limited. We do not yet know when, where or how the first magnetic fields in the Universe were formed, nor do we fully understand their role in fundamental processes such as galaxy formation or cosmic ray acceleration or how they influence the evolution of astrophysical objects. The greatest challenge to addressing these issues has been a lack of deep, broad bandwidth polarimetric data over large areas of the sky. The Square Kilometre Array will radically improve this situation via an all-sky polarisation survey that delivers both high quality polarisation imaging in combination with observations of 7-14 million extragalactic rotation measures. Here we summarise how this survey will improve our understanding of a range of astrophysical phenomena on scales from individual Galactic objects to the cosmic web.

astro-ph.CO↗

arXiv2015

SKA Deep Polarization and Cosmic Magnetism

Deep surveys with the SKA1-MID array offer for the first time the opportunity to systematically explore the polarization properties of the microJy source population. Our knowledge of the polarized sky approaching these levels is still very limited. In total intensity the population will be dominated by star-forming and normal galaxies to intermediate redshifts ($z \sim1-2$), and low-luminosity AGN to high redshift. The polarized emission from these objects is a powerful probe of their intrinsic magnetic fields and of their magnetic environments. For redshift of order 1 and above the broad bandwidth of the mid-bands span the Faraday thick and thin regimes allowing study of the intrinsic polarization properties of these objects as well as depolarization from embedded and foreground plasmas. The deep field polarization images will provide Rotation Measures data with very high solid angle density allowing a sensitive statistical analysis of the angular variation of RM on critical arc-minute scales from a magnetic component of Large Scale Structure of the Universe.

astro-ph.GA↗

arXiv2015

Magnetic Field Tomography in Nearby Galaxies with the Square Kilometre Array

Magnetic fields play an important role in shaping the structure and evolution of the interstellar medium (ISM) of galaxies, but the details of this relationship remain unclear. With SKA1, the 3D structure of galactic magnetic fields and its connection to star formation will be revealed. A highly sensitive probe of the internal structure of the magnetoionized ISM is the partial depolarization of synchrotron radiation from inside the volume. Different configurations of magnetic field and ionized gas within the resolution element of the telescope lead to frequency-dependent changes in the observed degree of polarization. The results of spectro-polarimetric observations are tied to physical structure in the ISM through comparison with detailed modeling, supplemented with the use of new analysis techniques that are being actively developed and studied within the community such as Rotation Measure Synthesis. The SKA will enable this field to come into its own and begin the study of the detailed structure of the magnetized ISM in a sample of nearby galaxies, thanks to its extraordinary wideband capabilities coupled with the combination of excellent surface brightness sensitivity and angular resolution.

astro-ph.GA↗