HEALFormer: AI Advances Weak Lensing Mass Mapping

This paper introduces HEALFormer, a novel transformer-based neural network architecture designed to enhance weak gravitational lensing mass mapping. By operating directly on the Hierarchical Equal Area isoLatitude Pixelization (HEALPix) and employing learnable mask tokens, HEALFormer effectively handles incomplete and noisy shear observations from various survey geometries. The model's progressive training strategy enables efficient processing of high-resolution maps, demonstrating excellent performance across diverse survey footprints, including KiDS, DES, DECaLS, and Planck. A key achievement of HEALFormer is its ability to surpass the theoretical phase recovery limits of linear reconstruction methods at small scales, representing a significant breakthrough in weak lensing analysis. The model's robust generalization to cosmological parameters beyond its training set and superior noise suppression capabilities further solidify its value for current and next-generation cosmological surveys. The availability of the code on GitHub promotes reproducibility and facilitates further development within the research community. HEALFormer's computational efficiency, reconstruction accuracy, and adaptability to varying survey configurations position it as a promising tool for advancing our understanding of dark matter distribution and testing cosmological models.

_Context: This intelligence report was compiled by the DailyOrbitalWire Strategy Engine. Verified for Art. 50 Compliance._