AS-Bridge: A Bidirectional Generative Framework Bridging Next-Generation Astronomical Surveys CCS: Applied computing Astronomy CCS: Computing methodologies Reconstruction CCS: Computing methodologies Anomaly detection

Dichang Zhang, Yixuan Shao, Simon Birrer, Dimitris Samaras · 2026

Read AS-Bridge: A Bidirectional Generative Framework Bridging Next-Generation Astronomical Surveys CCS: Applied computing Astronomy CCS: Computing methodologies Reconstruction CCS: Computing methodologies Anomaly detection by Dichang Zhang, Yixuan Shao, Simon Birrer, Dimitris Samaras (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-11928.txt.

Introduction

Zhang and co-authors introduce AS-Bridge, a generative model translating between the LSST and Euclid sky surveys. The ground-based LSST and the space-based Euclid differ in depth, wavelength, and coverage, making joint work rich but hard. AS-Bridge learns a diffusion model on a Brownian Bridge path between the two views. Where their skies overlap, it models how one view predicts the other. This enables faithful guesses of missing observations and hunts for rare events. It shows survey-to-survey modeling is feasible.

Catalog record

CREATORDichang Zhang, Yixuan Shao, Simon Birrer, Dimitris Samaras
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46392
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2603.11928

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

AS-Bridge: A Bidirectional Generative Framework Bridging Next-Generation Astronomical Surveys CCS: Applied computing Astronomy CCS: Computing methodologies Reconstruction CCS: Computing methodologies Anomaly detection Dichang Zhang, Yixuan Shao, Simon Birrer, Dimitris Samaras Abstract . The upcoming decade of observational cosmology will be shaped by large sky surveys, such as the ground-based LSST at the Vera C. Rubin Observatory and the space-based Euclid mission. While they promise an unprecedented view of the Universe across depth, resolution, and wavelength, their differences in observational modality, sky coverage, point-spread function, and scanning cadence make joint analysis beneficial, but also challenging. To facilitate joint analysis, we introduce A(stronomical)S(urvey)-Bridge, a bidirectional generative model that translates between ground- and space-based observations. AS-Bridge learns a diffusion model that employs a stochastic Brownian Bridge process between the LSST and Euclid observations. The two surveys have overlapping sky regions, where we can explicitly model the conditional probabilistic distribution between them. We show that this formulation enables new scientific capabilities beyond single-survey analysis, including faithful probabilistic predictions of missing survey observations and inter-survey detection of rare events. These results establish the feasibility of inter-survey generative modeling. AS-Bridge is therefore well-positioned to serve as a complementary component of future LSST–Euclid joint data pipelines, enhancing the scientific return once data from both surveys become available. Data and code are available at https://github.com/ZHANG7DC/AS-Bridge . Figure 1. Overview of AS-Bridge. The central panel shows the visible sky, with the LSST and Euclid survey footprints marked in blue and red, respectively. The overlapping regions indicate areas jointly observed by the ground-based LSST (left) and the space-based Euclid mission (right). LSST provides multi-band optical images that are more blended due to atmospheric seeing, while Euclid delivers sharper near-infrared observations from space. From these overlapping regions, matched image cutouts are extracted and used to train AS-Bridge, which models the probabilistic translation between the two survey domains using a Brownian bridge formulation. 1. Introduction In this decade, observational cosmology will be driven by major flagship surveys, such as the ground-based Legacy Survey of Space and Time (LSST) at the NSF–DOE Vera C. Rubin Observatory ( 19 ) and the space-based ESA Euclid mission ( 10 ) . Together, they provide unprecedented coverage across depth, angular resolution, wavelength, and temporal cadence, enabling major advances in topics ranging from dark energy and dark matter to asteroids and other solar system objects ( 5 ; 4 ; 19 ; 31 ; 33 ; 22 ) . Fundamental differences in observational modality—such as Point Spread Functions (PSFs), noise statistics, photometric bandpasses, and scanning cadence—introduce systematic distribution shifts that complicate cross-survey inference. The mapping between ground-based and space-based imaging is inherently ill-posed in both directions. Recovering Euclid-quality morphology from LSST requires resolving ambiguity induced by atmospheric blurring and background noise. Conversely, mapping Euclid observations back to the LSST domain requires inferring spectral information from fewer and different bands, making the relationship non-identifiable from a single deterministic mapping. Therefore, inter-survey translation should be treated as a probabilistic process that can sample multiple valid realizations consistent with the available observations. To model this probabilistic process, we introduce AS-Bridge, a unified bidirectional generative framework designed to model the stochastic connection between survey domains. AS-Bridge formulates inter-survey translation as a bidirectional Brownian bridge process, learning a stochastic path that connects the LSST and Euclid data distributions using overlapping sky regions as anchors. This formulation explicitly models the probabilistic transition between the two surveys and enables calibrated probabilistic inference in both directions. We demonstrate the following benefits from inter-survey generation: Inter-survey restoration and synthesis. After training on paired cutouts—currently from simulations, but which can be readily constructed once official data releases become available after camera calibration—AS-Bridge can generate complementary observations in non-overlapping sky regions. We show that AS-Bridge recovers the target-survey observation conditioned on the source-survey input, and empirically demonstrate superior probabilistic reconstruction quality compared to competing baselines (Sec. 4.3.2 ). This enables analyses developed for one survey to be transferred to regions observed only by the other, while preserving uncertainty quantification. Inter-survey rare event detection. We cast rare-event detection as inter-survey inconsistency under uncertainty. At inference time, a paired observation is mapped into the inter-survey bridge and reconstructed back to one domain; the reconstruction inconsistency, aggregated over multiple stochastic samples, serves as an anomaly score. This yields an anomaly detector that simultaneously exploits complementary observational modalities through cross-survey conditioning and probabilistic reconstruction along the diffusion bridge. We demonstrate improved detection of strong gravitational lenses as a representative example of structural outliers, compared to general-purpose baselines. Our contributions are as follows: • We introduce the inter-survey modeling problem and formulate probabilistic inter-survey inference for next-generation cosmological surveys by treating cross-survey translation as an ill-posed, bidirectional inference problem. We further identify rare event detection as a key downstream application enabled by this formulation. • We propose AS-Bridge, a generative framework that models inter-survey translation as a bidirectional Brownian bridge process, recovers the target conditional distribution given the source survey observation, and enables rare event detection through inter-survey probabilistic inconsistency. • We construct the first realistic simulated benchmark for this problem and systematically compare AS-Bridge against representative baseline methods under standardized evaluation protocols and metrics. 2. Background and Related Work 2.1. Next-Generation Astronomical Surveys This decade marks a paradigm shift in observational cosmology driven by the emergence of wide-field, high-throughput sky surveys. These next-generation surveys aim to map the Universe with unprecedented depth, angular resolution, spectral coverage, and temporal cadence, enabling precision studies of cosmology, galaxy evolution, and time-domain phenomena at scale. Among them, the ground-based Legacy Survey of Space and Time (LSST) ( 19 ) at the Vera C. Rubin Observatory, the space-based Euclid mission, the Nancy Grace Roman Observatory ( 26 ) , the Square Kilometer Array (SKA) ( 1 ) and the Simon’s Observatory. In this paper we focus on LSST and Euclid, as these are the first surveys on the sky with high complementary in resolution and temporal cadence. LSST is a deep, multi-band optical imaging survey conducted by the Vera C. Rubin Observatory, located on Cerro Pachón in northern Chile. Its southern-hemisphere location enables continuous access to a large, contiguous region of the southern sky. Over a planned ten-year survey, LSST will repeatedly image approximately 18,000 deg 2 of sky across six optical bands ( u,g,r,i,z,y ), with a typical revisit cadence of a few days. The telescope features an effective aperture of 6.7 m and a 9.6 deg 2 field of view, producing images with a typical seeing-limited angular resolution of \sim 0.7 arcsec. The strengths of LSST lie in its exceptional depth, temporal coverage, and statistical power, making it particularly well suited for studies of large-scale structure, weak gravitational lensing, and transient and variable phenomena. However, as a ground-based survey, LSST observations are affected by atmospheric turbulence, variable seeing conditions, sky background, and weather-dependent systematics. In contrast, Euclid ( 10 ) is a space-based mission operated by the European Space Agency, designed to deliver high-resolution imaging and near-infrared photometry and spectroscopy over a wide area of the extragalactic sky. Free from atmospheric effects, Euclid provides stable and sharply defined point-spread functions, enabling precise galaxy shape measurements. Euclid carries two primary instruments: the VIS instrument, which provides high-resolution optical imaging in a broad visible band with a

Read the complete text