Strong Gravitational Lensing Posterior Sampling in Pixel-Space Using Diffusion Models and Recurrent Inference Machines

2026-07-21

AI Summary

This work addresses the computational bottleneck in jointly inferring the source galaxy surface brightness and foreground mass distribution from strong gravitational lensing images under high-dimensional, nonlinear conditions. The authors propose a novel approach that, for the first time, integrates diffusion generative models with a recursive inference mechanism to efficiently sample the joint posterior distribution directly in pixel space. Leveraging high-fidelity lensing data generated from cosmological hydrodynamical simulations, the method achieves reconstructions of both the source galaxy and mass distribution at noise-level precision, substantially overcoming the limitations of traditional techniques in handling high dimensionality and nonlinearities.

Abstract

Modeling galaxy-galaxy strong gravitational lenses to infer the brightness of the source galaxy and the mass distribution of the foreground galaxy is computationally challenging, particularly for high-resolution, high signal-to-noise ratio observations. In this regime, high-dimensional representations of both the source and the foreground mass distribution are necessary to model the data down to the noise level. This inference problem has been challenging for both traditional and machine learning-based methods because of its high dimensionality and its non-linearity in the foreground mass distribution. We present a method to generate joint posterior samples of the source galaxy and foreground mass distribution as pixelated images conditioned on observations. The method combines diffusion-based generative modeling and recurrent inference machines. It can model realistic gravitational lensing simulations with background and foreground galaxies drawn from cosmological hydrodynamical simulations down to the noise level.

Problem

• strong gravitational lensing
• posterior sampling
• high-dimensional inference
• mass distribution
• source reconstruction

Innovation

• diffusion models
• recurrent inference machines
• strong gravitational lensing
• posterior sampling
• pixel-space inference