B. Jin | Searcharxiv

arXiv2026

MIGHTEE/COSMOS-3D: The discovery of three spectroscopically confirmed radio-selected star-forming galaxies at z=4.9-5.6

Radio observations offer a dust-independent probe of star formation and active galactic nucleus (AGN) activity, but sufficiently deep data are required to access the crossover luminosity between these processes at high redshift ($z>4.5$). We present three spectroscopically confirmed high-redshift radio sources (HzRSs) detected at 1.3 GHz at $z=4.9$-$5.6$, with radio luminosities spanning $L_{\rm 1.3 \, GHz}\approx2$-$5\times10^{24} \, \rm W \, Hz^{-1}$. These sources were first identified as high-redshift candidates through spectral energy distribution (SED) fitting of archival Hubble, JWST NIRCam+MIRI, and ground-based photometry, and then spectroscopically confirmed via the $\rm H\,\alpha$ emission line using wide-field slitless spectroscopy from JWST COSMOS-3D. The star formation rates (SFRs) measured from SED fitting, the $\rm H\,\alpha$ flux, and the 1.3 GHz luminosity, span $\sim100$-$1800\, M_{\odot} \, \rm yr^{-1}$, demonstrating broad agreement between these SFR tracers. We find that these three sources lie either on or $0.5$-1.0 dex above the star-forming main sequence at $z=4$-6 and have undergone a recent burst of star formation. The sources have extended rest-UV/optical morphologies with no evidence for a dominant point source component, indicating that an AGN is unlikely to dominate their rest-UV and optical emission. Two of the sources have complex, multi-component rest-frame UV/optical morphologies, suggesting that their starbursts may be triggered by merging activity. These HzRSs open up a new window towards probing radio emission powered by star formation alone at $z> 4.5$, representing a remarkable opportunity to begin tracing star formation, independent of dust, in the early Universe.

astro-ph.GA↗

arXiv2023

On the Convergence of Stochastic Gradient Descent for Linear Inverse Problems in Banach Spaces

In this work we consider stochastic gradient descent (SGD) for solving linear inverse problems in Banach spaces. SGD and its variants have been established as one of the most successful optimisation methods in machine learning, imaging and signal processing, etc. At each iteration SGD uses a single datum, or a small subset of data, resulting in highly scalable methods that are very attractive for large-scale inverse problems. Nonetheless, the theoretical analysis of SGD-based approaches for inverse problems has thus far been largely limited to Euclidean and Hilbert spaces. In this work we present a novel convergence analysis of SGD for linear inverse problems in general Banach spaces: we show the almost sure convergence of the iterates to the minimum norm solution and establish the regularising property for suitable a priori stopping criteria. Numerical results are also presented to illustrate features of the approach.

cs.LG↗