Search Talks

Lambda Cold Dark Matter (LCDM) is the most successful theory for the formation of structure in the Universe. Although its predictions have been verified on large scales, they are still contested on the scale of dwarf galaxies, whose dynamical properties are often cited as evidence for the need to revise some of LCDM’s basic tenets. In this context, I will discuss the recent discovery of the faintest galaxies known to date, and how their properties may be used to place constraints on the clustering of dark matter on the smallest galactic and sub-galactic scales, as well as on the viability of some of the proposed alternatives to LCDM.

Modern telescopes such as Hubble and JWST deliver exquisitely detailed images that encode rich information about galaxies, stars, and dark matter. Extracting this information requires solving high-dimensional inverse problems in which each pixel represents a parameter of the sky. In this talk, I will present a Bayesian framework that uses diffusion-based generative models to perform inference directly in these high-dimensional spaces. By learning expressive priors over galaxy morphologies and likelihoods over complex, non-Gaussian noise distributions, this approach enables the reconstruction of astronomical images at unprecedented fidelity. This talk will cover a wide range of applications where we have applied these techniques, from the recovery of detailed images of protoplanetary discs using ALMA data and high redshift galaxies using Hubble data and strong gravitational lensing to a discussion about the misscalibration of the generative model. The generative model often needs to be trained on low redshift galaxies or using simulations, all of which could introduce biases in our inference. To address this, I will present a framework which can update the parameters of the generative model using corrupted data only. Together, these results illustrate how generative modeling can be used in a principled statistical framework for astronomy.

Galaxy science and cosmology are rapidly extending to ever higher redshifts, where new observations have motivated a wide range of galaxy formation models. Because these models are all tuned to reproduce observed number densities, clustering provides the critical observable for breaking their degeneracies. Pure-parallel surveys, often underexploited, enable clustering to be measured through field-to-field variance across a broad redshift range. I will show measurements from the JWST PANORAMIC survey, spanning quiescent galaxies at intermediate redshift to luminous Lyman-break galaxies above z∼10. Linear-bias results from COSMOS-Web up to z∼8 reveal a consistent picture: galaxies in the early universe are highly clustered, placing strong constraints on models of early formation. Remarkably, even carefully tuned models struggle to reproduce the observed amplitudes. Cosmic variance also emerges as both a systematic and a tool in number-count statistics, with particular relevance for explaining the presence of ultra-massive quiescent galaxies already in the young universe. I will also provide a brief overview of complementary low-redshift work, including the use of small-scale clustering to reconstruct halo merger trees with graph neural networks, and other efforts to maximize the scientific return of the next generation of galaxy surveys.