Blind Strong Gravitational Lensing Inversion: Joint Inference of Source and Lens Mass with Score-Based Models

2025-11-06

AI Summary

In strong gravitational lensing inversion, conventional methods rely on predefined parametric lens mass models, hindering unbiased joint inference of the source galaxy and lens parameters. This paper introduces score-based generative modeling to blind strong lens inversion for the first time, proposing a prior-free framework for joint posterior inference. Our method employs the GibbsDDRM continuous-time sampler to jointly optimize high-dimensional image space (source galaxy) and low-dimensional parameter space (lens mass distribution), while leveraging a data-driven score prior to regularize source structure. Experiments demonstrate that reconstruction residuals are statistically consistent with observational noise, and marginal posteriors of lens parameters accurately converge to ground-truth values. The approach exhibits both effectiveness and robustness under realistic astronomical conditions, enabling model-agnostic, end-to-end probabilistic inference without assuming a specific lens mass profile.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Score-based models can serve as expressive, data-driven priors for scientific inverse problems. In strong gravitational lensing, they enable posterior inference of a background galaxy from its distorted, multiply-imaged observation. Previous work, however, assumes that the lens mass distribution (and thus the forward operator) is known. We relax this assumption by jointly inferring the source and a parametric lens-mass profile, using a sampler based on GibbsDDRM but operating in continuous time. The resulting reconstructions yield residuals consistent with the observational noise, and the marginal posteriors of the lens parameters recover true values without systematic bias. To our knowledge, this is the first successful demonstration of joint source-and-lens inference with a score-based prior.

Problem

• Jointly inferring source galaxy and parametric lens mass distribution
• Overcoming previous assumption of known lens mass in gravitational lensing
• Using score-based models as expressive priors for inverse problems

Innovation

• Joint source and lens inference using score-based models
• GibbsDDRM sampler operating in continuous time
• Parametric lens-mass profile reconstruction without bias