The next generation of radio astronomy telescopes are challenging existing data analysis paradigms, as they have an order of magnitude more antennas and larger bandwidth. Foremost amongst these are deep spectral line surveys, because these have the largest number of epochs and spectral channels per dataset. For example, the Deep Investigation of Neutral Gas Origins (DINGO) project on the Australian Square Kilometre Array Pathfinder (ASKAP) aims to observe over 3,200 hours spread over hundreds of observing sessions, covering two tiles, two footprints and two frequency settings. The two primary problems encountered when processing this data are the need for storage and that processing is primarily I/O limited. To address these issues, we have implemented a deep imaging pipeline based on the storage of an intermediate data product in the software ASKAPsoft, that of the uv-gridded data, and have demonstrated lossy and lossless compression of this data on ASKAP, using MGARD and ADIOS2 libraries. We find data compression ratios from a factor of 7 (lossless) up to 20 (using lossy compression with an absolute error bound of $10^{-4}$), and processing is significantly faster for lossless compression. We discuss the effectiveness of lossy MGARD compression and its adherence to the designated error bounds, the trade-off between these error bounds and the corresponding compression ratios, as well as the potential consequences of these I/O and storage improvements on the science quality of the data products. As lossless compression allows us to achieve the DINGO goals within the storage limitations for the project, this will be the option adopted.

We present SM-Net, a machine learning model that learns a continuous spectral manifold from multiple high-resolution stellar libraries. SM-Net generates stellar spectra directly from fundamental stellar parameters such as effective temperature ($T_{\mathrm{eff}}$), surface gravity (${\log}\, g \equiv {\log}_{10}[g/(\mathrm{cm}\,\mathrm{s}{^{-2}})]$), and metallicity ($\log Z \equiv \log_{10}\kern-1.5pt(Z/Z_{\odot})$), trained on a combined grid derived from PHOENIX–Husser, C3K–Conroy, OB–PoWR, and TMAP–Werner libraries. By combining their respective parameter spaces to achieve broader coverage, we construct an extended composite dataset that spans a significantly larger, more continuous region of stellar parameter space than any single library. The unified grid spans $T_{\mathrm{eff}} = 2{\,}000$–$190{\,}000{\,}$K, $\log g = -1$ to 9, and $\log Z = -4$ to 1 dex, with spectra covering the wavelength range 3 000–100 000 Å. Within the combined parameter space, SM-Net provides smooth interpolation across heterogeneous library boundaries. Outside the sampled domain, it can generate numerically smooth exploratory predictions, although these extrapolated spectra are not directly validated against reference models. Zero or masked flux values are treated as unknowns rather than physical zeros, allowing the network to infer missing regions using correlations learned from neighbouring grid points. When evaluated over the unique, unaugmented splits of 3 526 training and 11 509 test spectra, SM-Net achieves mean per-spectrum mean-squared error values of $1.03\times10^{-5}$ and $1.97\times10^{-5}$, respectively, in the transformed log1p-scaled flux representation. These results confirm that the learned manifold reproduces the spectra with high accuracy while remaining robust to incomplete or masked data. Inference throughput exceeds $14\,000$ spectra per second on a single graphics processing unit. This throughput enables rapid exploration of the stellar parameter space and efficient construction of synthetic stellar population models. To promote accessibility and community use, we make the SM-Net model publicly available, along with an interactive dashboard that enables real-time spectral generation and visualisation through a web interface. Users can explore the influence of $T_{\mathrm{eff}}$, $\log g$, and $\log Z$ interactively, thereby demonstrating a deployed research model ready for exploration. SM-Net is a fast, robust, and flexible data-driven complement to traditional stellar population synthesis libraries, providing a pathway for next-generation stellar population modelling.

Wind-fed accretion onto magnetised neutron stars drives long-term spin variability that can be used to trace angular-momentum transfer in structured stellar outflows. We analyse five years of TESS optical timing to measure the spin evolution of the wind-fed X-ray pulsar 4U 1954+319, which hosts the slowest known accretion-powered pulsar. Coherent hour-scale pulsations are detected in all eight TESS sectors. The optical spin period evolves non-monotonically from $5.51\pm0.06$ to $5.72\pm0.06$ h and then returns to $5.55\pm0.05$ h, consistent with a mid-2023 break and a torque reversal from spin-down to spin-up ($\kern1.5pt\dot{\!P}_1=+0.68\pm0.08$ s d$^{-1}$ to $\kern1.5pt\dot{\!P}_2=-2.16\pm0.35$ s d$^{-1}$). The observed $\kern1.5pt\dot{\!P}$ corresponds to torques of $N_1=(\!-\!1.2\pm0.1)\times10^{32}$ dyn cm and $N_2=(\!+\!3.6\pm0.6)\times10^{32}$ dyn cm for a canonical moment of inertia $I=10^{45}$ g cm$^2$, while the associated spin-down power remains negligible compared with the X-ray luminosity. Within the quasi-spherical settling accretion regime, a change in mass accretion rate by a factor of ${\sim}3.6$ is sufficient, at the order-of-magnitude level, to account for the measured torque ratio. The inferred torque scale is compatible with a surface dipole field $B\approx(0.5-5)\times10^{12}$ G, with the magnetospheric radius remaining well inside co-rotation and no requirement for a propeller transition at the inferred luminosities. No clear contemporaneous hard-X-ray brightening is apparent in the BAT monitoring, which is consistent with an origin of the optical pulsations in reprocessing of the pulsed high-energy emission. Overall, the data favour a picture in which 4U 1954+319 is a long-period, near-equilibrium wind-fed pulsar whose torque reversals can arise from moderate changes in settling accretion, without invoking magnetar-strength fields or persistent propeller transitions. They also show that long-baseline optical timing can provide useful constraints on torque balance in faint wind-fed pulsars.

Obtaining robust galaxy number counts is crucial for understanding galaxy evolution, and submillimetre counts in particular have proven valuable for revising subgrid physics models in cosmological simulations. In confusion-limited surveys, which are common at these wavelengths, statistical methods such as P(D) fluctuation analysis are required to recover counts of faint, unresolved galaxies. However, the standard P(D) framework assumes that galaxies are Poisson-distributed, whereas in reality galaxies are clustered. Using simulations, we demonstrate that this clustering systematically biases P(D)-derived number counts and present an empirical method that simultaneously measures and corrects for this bias by combining the 1- and 2-point statistics in the map, thereby maximising the information extracted from the data. Applying this method to deep Herschel-SPIRE observations of the GOODS-N field, we provide revised galaxy number counts at 250, 350, and 500 $\unicode{x03BC}$m. Our results indicate that at 500 $\unicode{x03BC}$m clustering inflates the apparent counts by a factor of 1.6 around $10\,$mJy and slightly suppresses the faintest sub-mJy counts, with milder effects at 350 and 250 $\unicode{x03BC}$m owing to the smaller beam sizes. This methodology is broadly applicable to other confusion-limited datasets with well-characterised beam and noise properties, including Spitzer, SCUBA-2, and the upcoming CCAT, enabling unbiased exploitation of the full statistical information in many surveys in the mid-IR to radio wavebands.

Understanding the evolution of the cosmic star formation rate density (SFRD) is key to uncovering how the Universe arrived at its present state. This paper presents a novel and efficient method to estimate the local SFRD, which uses supervised machine learning to first identify a population of star-forming galaxies (SFGs). Next, star-formation rates (SFRs) are determined using the 1.4-GHz radio-continuum emission detected by the Australian Square Kilometre Array Pathfinder (ASKAP). Specifically, a gradient-boosted decision tree model was implemented to classify extragalactic sources from the Beck et al. (2022, MNRAS, 515, 4711) catalogue as either galaxies or quasars using RACS-mid and WISE photometry. The full sample, consisting of 389 392 sources, was partitioned into a 70%–15%–15% split for training, validating, and testing. The optimised model achieved a weighted F1 score of 0.93 and an accuracy of 0.94 on the test dataset, ultimately classifying 336 674 sources as galaxies and 52 718 sources as quasars. Using the resulting $z\lt0.1$ depth-matched galaxy sample and the photometric redshift predictions from Beck et al. (2022, MNRAS, 515, 4711), a modified 1.4-GHz SFR calibration was determined, yielding a local, completeness-corrected, $z\lt0.1$ SFRD of $(1.4 \pm 0.5) \times 10^{-2} \; \rm M_{\odot}\, yr^{-1}\,Mpc^{-3}$ using 11 293 sources. This value is consistent with previous results. Thus, this study demonstrates the feasibility of using supervised learning to identify large populations of SFGs in order to investigate the SFRD evolution. This presents an exciting prospect for future, deeper surveys such as EMU, which will enable the cosmic SFRD to be probed out to higher redshifts.

Overlapping galaxies, in which a foreground galaxy partially overlaps a background galaxy, offer a unique opportunity to measure dust attenuation, a key nuisance parameter in galaxy studies, empirically and in great detail by modelling the light of both the foreground and background galaxy and inferring the missing light in the overlapping region. However, the current catalogue of overlapping pairs is relatively limited in number compared to catalogues dedicated to individual galaxies. Expanding this catalogue is not only a necessity to facilitate further detailed dust studies beyond the few limited studies conducted thus far but also to improve pair-to-pair variance and support automated identification through machine learning techniques. To achieve this, we utilise galaxies classified as ‘overlapping’ from Galaxy Zoo DECaLS (GZD-1, -2, and -5), along with images from Data Release 10 (DR10) of the DESI Legacy Survey, in our individual citizen science project to classify these pairs directly using volunteers. This new catalogue will not only provide a wealth of targets for future dust studies but will also contribute to a deeper understanding of these pairs and dust as a whole.

We present independent imaging analyses of Event Horizon Telescope (EHT) observations of the active galactic nuclei in radio galaxy Centaurus A and quasar 3C 279 using Generative Deep learning Image Reconstruction with Closure Terms (GenDIReCT), a recently developed machine-learning framework built on conditional diffusion models that uses interferometric closure invariants as primary observables. For Centaurus A, our reconstruction reveals two prominent emission ridges ($\simeq 80\,\unicode{x03BC}$as each) along the jet sheath with a brightness ratio of $1.4\pm 0.1$ and an opening angle of $12.3\pm 0.3$ deg. For 3C 279, we identify three distinct components in the image, with the southern jet ejecta on sub-parsec scale exhibiting a proper motion of $4.6\pm 1.0\,\unicode{x03BC}$as over $\approx 5.39$ d away from the northern components, corresponding to an apparent superluminal velocity of $\simeq 10\pm 2$ times light speed. These measurements are consistent with those reported by the EHT Collaboration. The results are significant because we demonstrate that: (1) imaging from interferometric aperture synthesis data, especially in VLBI and most acutely in extremely sparse arrays like the EHT, remains a severely ill-posed and challenging inverse problem, yet closure invariants preserve robust morphological information that can strongly constrain structural features, and (2) more importantly, closure-invariant imaging largely avoids calibration systematics, thus providing a fundamentally independent view of spatial structure with very high angular resolution. The generative nature of GenDIReCT further allows us to sample and characterise clusters of plausible image solutions for each dataset. As a calibration-independent, generative imaging approach, GenDIReCT offers a robust and truly independent blind-imaging tool for current and future VLBI experiments.

With the installation of next-generation phased array feed (PAF) receivers on radio telescopes, there is an urgent need to develop effective and computationally efficient radio frequency interference (RFI) mitigation methods for large-scale surveys. Here, we present a new RFI mitigation package, called mRAID (multi-beam RAdio frequency Interference Detector), which uses the eigenvalue decomposition algorithm to identify RFI in cross-correlation matrix (CCM) of data recorded by multiple beams. When applied to high time-resolution pulsar search data from the Five-hundred-meter Aperture Spherical Radio Telescope (FAST), mRAID demonstrates excellent performance in identifying RFI over short timescales, thereby enhancing the efficiency of pulsar and fast radio burst (FRB) searches. Since the computation of the CCM and the eigenvalue decomposition for each time sub-integration and frequency channel are independent, the process is fully parallelisable. As a result, mRAID offers a significant computational advantage over commonly used RFI detection methods.

Bars are ubiquitous morphological features in the observed distribution of galaxies. There are similarly many methods for classifying these features and, without a strict theoretical definition or common standard practice, this is often left to circumstance. So, we were concerned whether astronomers even agree on the bar which they perceive in a given galaxy and whether this could impact perceived scientific results. As an elementary test, we twenty-one astronomers with varied experience in studying resolved galaxies and circumstances, have each assessed 200 galaxy images, spanning the early phase of bar evolution in two different barred galaxy simulations. We find variations exist within the classification of all the standard bar parameters assessed: bar length, axis-ratio, pitch-angle and even whether a bar is present at all. If this is indicative of the wider community, it has implications for interpreting morphological trends, such as bar-end effects. Furthermore, we find that it is surprisingly not expertise but gender, followed by career stage, which gives rise to the largest discrepancies in the reported bar parameters. Currently, automation does not seem to be a viable solution, with bar classifications from two automated bar-finding algorithms tested and failing to find bars in snapshots where most astronomers agree a bar must exist. Increasing dependence on machine learning or crowdsourcing with a training dataset can only serve to obfuscate any existing biases if these originate from the specific astronomer producing the training material. On the strength of this small sample, we encourage an interim best practice to reduce the impact of any possible classification bias and set goals for the community to resolve the issue in the future.

Image reconstruction in very-long baseline interferometry operates under severely sparse aperture coverage with calibration challenges from both the participating instruments and propagation medium, which introduce the risk of biases and artefacts. Interferometric closure invariants offer calibration-independent information on the true source morphology, but the inverse transformation from closure invariants to the source intensity distribution is an ill-posed problem. In this work, we present a generative deep learning approach to tackle the inverse problem of directly reconstructing images from their observed closure invariants. Trained in a supervised manner with simple shapes and the CIFAR-10 dataset, the resulting trained model achieves reduced chi-square data adherence scores of $\chi^2_{\mathrm{CI}} \lesssim 1$ and maximum normalised cross-correlation image fidelity scores of $\rho_{\mathrm{NX}} \gt 0.9$ on tests of both trained and untrained morphologies, where $\rho_{\mathrm{NX}}=1$ denotes a perfect reconstruction. We also adapt our model for the Next Generation Event Horizon Telescope total intensity analysis challenge. Our results on quantitative metrics are competitive to other state-of-the-art image reconstruction algorithms. As an algorithm that does not require finely hand-tuned hyperparameters, this method offers a relatively simple and reproducible calibration-independent imaging solution for very-long baseline interferometry, which ultimately enhances the reliability of sparse very-long baseline radio interferometry imaging results.

The Hector Galaxy Survey is a new optical integral field spectroscopy (IFS) survey currently using the Anglo-Australian Telescope to observe up to 15 000 galaxies at low redshift ($z \lt 0.1$). The Hector instrument employs 21 optical fibre bundles feeding into two double-beam spectrographs, AAOmega and the new Spector spectrograph, to enable wide-field multi-object IFS observations of galaxies. To efficiently process the survey data, we adopt the data reduction pipeline developed for the SAMI Galaxy Survey, with significant updates to accommodate Hector’s dual-spectrograph system. These enhancements address key differences in spectral resolution and other instrumental characteristics relative to SAMI and are specifically optimised for Hector’s unique configuration. We introduce a two-dimensional arc fitting approach that reduces the root-mean-square (RMS) velocity scatter by a factor of 1.2–3.4 compared to fitting arc lines independently for each fibre. The pipeline also incorporates detailed modelling of chromatic optical distortion in the wide-field corrector, to account for wavelength-dependent spatial shifts across the focal plane. We assess data quality through a series of validation tests, including wavelength solution accuracy (1.2–2.7 km s$^{-1}$ RMS), spectral resolution (FWHM of 1.2–1.4 Å for Spector), throughput characterisation, astrometric precision ($\lesssim$ 0.03 arcsec median offset), sky subtraction residuals (1–1.6% median continuum residual), and flux calibration stability (4% systematic offset when compared to Legacy Survey fluxes). We demonstrate that Hector delivers high-fidelity, science-ready datasets, supporting robust measurements of galaxy kinematics, stellar populations, and emission-line properties and provide examples. Additionally, we address systematic uncertainties identified during the data processing and propose future improvements to enhance the precision and reliability of upcoming data releases. This work establishes a robust data reduction framework for Hector, delivering high-quality data products that support a broad range of extragalactic studies.

We investigate whether the SOL2024-04-17T22:10 (M1.0) flare in AR 13645 was accompanied by a coronal mass ejection (CME) by jointly analyzing EUV imagery from SDO/AIA and white-light observations from SOHO/LASCO C2. Using GOES/XRS as a timing reference, we examine AIA 171 and 131 Å sequences before, during, and after the flare and identify post-flare expansion of overlying loop systems without signatures of filament eruption. Running-difference LASCO/C2 frames reveal a faint, partial-halo front between 03:12:07 and 03:24:07 UT on 2024-04-18 propagating in the quadrant of AR 13645, with a cataloged plane-of-sky speed of $\sim$246 km s$^{-1}$, position angle $\approx$238$^\circ$, and angular width $\approx$248$^\circ$. Co-registration in JHelioviewer shows temporal and geometrical consistency between the AIA loop expansion and the LASCO front. We therefore interpret the event as a low-contrast, low-energy CME possibly associated with the flare, while noting the limitations imposed by the faintness of the white-light signal and the hours-long delay between the flare peak and the coronagraph detection. The analysis establishes observational constraints and a simple EUV-white-light workflow that could be extended with DEM estimates, CME mass/energy, and high-energy diagnostics in future works.

We analyze solar flares from 1976–2025 using the GOES/XRS_B event record to construct two monthly products: (i) counts of flares with peak flux $\ge$ B1 and (ii) mean XRS_B flux, summarized as year$\times$month heat-maps that separate productivity from energetics. We then pursue two complementary forecasting tracks. (A) Annualized SARIMAX: applying $\text{SARIMAX}(0,0,3)\times(1,1,1)_{11}$ to an annual mean series derived from daily counts improves over an annual naïve seasonal baseline (MSE $=2.37$ vs. $4.93$; MAE $=1.25$ vs. $1.80$), with diagnostics indicating near-normal, whitened residuals. (B) Monthly models: for monthly counts, simple Linear Regression, XGBoost, and LightGBM yield qualitatively reasonable envelopes but do not outperform a monthly naïve baseline; the naïve model attains RMSE $=70.06$ (MAE $=59$, MSE $=4909$), versus 104–207 RMSE for the ML baselines. The workflow delivers visual analytics and forecasts tied directly to GOES flare variables, with SARIMAX supporting cycle turning-point estimates and monthly models serving as cautious envelopes rather than precise long-horizon predictors.

We present a preliminary variability analysis of space-based photometry for 2,500 candidate pulsating stars, with physical parameters derived from Strömgren–Crawford photometry. Using an automatic classification pipeline, we identified and characterized different variability types and detected $\delta$ Scuti / $\gamma$ Doradus hybrid candidates, which will be further analyzed through detailed asteroseismic modeling of their frequency and period-spacing patterns.

Extended main-sequence turnoffs are observed in the color-magnitude diagrams of some young and intermediate-age star clusters. Here we report the case study of NGC 3532. Among the cluster member candidates identified using the ML-MOC algorithm on Gaia DR3 sources, we found that fast rotators are significantly redder. Stars near the main-sequence turnoff exhibit a multi-modal distribution in the Gaia broadening velocity ($v_{\rm broad}$). A comparison with relevant literature suggests binaries being typically interspersed with fast rotators, often manifesting as slow to moderate rotators. Additionally, chemically peculiar stars are predominantly located in the bluer regions of the turnoff, with those in the fast-rotator region having large values of Gaia re-normalised unit weight error, indicating plausible companions. Our analysis suggests multiple channels responsible for angular momentum loss to account for the extended turnoff.

This research presents unsupervised machine learning and statistical techniques to detect and analyze tidal tails in five open clusters, BH 164, Alessi 2, NGC 2281, NGC 2354, and M67, with ages ranging from 60 Myr to 4 Gyr, using Gaia DR3 data. Color-magnitude diagram (CMD) matching, principal component analysis (PCA), and density-based clustering were used to detect the tails. Tidal tails were found in all clusters, extending 40 to 100 pc and containing 100 to 200 stars. Other statistical methods were used to study the stellar kinematics, photometry, and spatial distribution of the detected features. The tails had luminosity functions similar to their parent clusters, generally lacked massive stars, and showed higher binary fractions. Significant rotation was detected in M67 and NGC 2281 for the first time.

The advent of next-generation radio telescopes is set to transform radio astronomy by producing massive data volumes that challenge traditional processing methods. Deep learning techniques have shown strong potential in automating radio analysis tasks, yet are often constrained by the limited availability of large annotated datasets. Recent progress in self-supervised learning has led to foundational radio vision models, but adapting them for new tasks typically requires coding expertise, limiting their accessibility to a broader astronomical community. Text-based AI interfaces offer a promising alternative by enabling task-specific queries and example-driven learning. In this context, large language models (LLMs), with their remarkable zero-shot capabilities, are increasingly used in scientific domains. However, deploying large-scale models remains resource-intensive, and there is a growing demand for AI systems that can reason over both visual and textual data in astronomical analysis. This study explores small-scale vision-language models (VLMs) as AI assistants for radio astronomy, combining LLM capabilities with vision transformers. We fine-tuned the LLaVA VLM on a dataset of 59k radio images from multiple surveys, enriched with 38k image-caption pairs from the literature. The fine-tuned models show clear improvements over base models in radio-specific tasks, achieving $\sim$30% F1-score gains in extended source detection, but they underperform vision-only classifiers and exhibit $\sim$20% drop on general multimodal tasks. Inclusion of caption data and LoRA fine-tuning enhances instruction following and helps recover $\sim$10% accuracy on multimodal benchmarks (e.g., ChartQA/DocVQA). This work lays the foundation for future advancements in radio VLMs, highlighting their potential and limitations, such as the need for better multimodal alignment, higher-quality datasets, and mitigation of catastrophic forgetting.

The Gaia mission has observed over 2 billion stars repeatedly across the entire sky over 10 years, revealing the many astronomical objects that vary on human timescales from seconds to years. Its repeated astrometric, photometric, spectrophotometric and spectroscopic measurements create an unprecedented dataset to probe the variable celestial sources down to $G\approx 21$ mag. To extract meaningful results from these many time series for so many sources, we have used machine learning techniques for crossmatching, variability detection, and variability classification. This approach has now led to the largest catalogue of classified variable sources ever produced over the entire celestial sphere.

We present a new framework for multi-line analysis that combines kernel principal component analysis (kernel PCA) and kernel shapley additive explanations (kernel SHAP). Applied to ALMA Band 3 data of NGC 1068, kernel PCA reveals features—including the AGN-driven outflow —not clearly captured by PCA, while kernel SHAP quantifies each line’s contribution. These results are consistent with previous studies and confirmed by a two-sample t-test. Moreover, kernel PCA scores correlate with LTE-derived column densities, highlighting the robustness of this framework.

We present a deep learning-based framework for automated quasar spectral analysis, addressing two key challenges: continuum estimation and absorption feature detection. Using autoencoders, convolutional neural networks (CNNs), and U-Net architectures, we demonstrate improved performance in reconstructing the quasar continuum and identifying absorption lines. These models are trained to learn complex, non-linear correlations across wide spectral regions, enabling robust predictions even in the presence of noise or missing data. Our methods offer greater flexibility and accuracy, particularly in low signal-to-noise regimes.

The U-Net architecture, in particular, shows strong potential for precise absorption feature detection, leveraging its encoder-decoder structure with skip connections to capture both local and global spectral patterns. This enables the identification of absorption signatures associated with metal-enriched gas in the circumgalactic medium. Our results highlight the scalability of these techniques for application to large-scale surveys such as DESI and WEAVE, making progress for new insights into the distribution and dynamics of baryons in the Universe.