analiza aiqbench
TNFlow applies transformer and normalizing flow architectures to invert synthetic radiative transfer models, enabling rapid inference of Trans-Neptunian Object surface compositions from JWST reflectance spectra. While the method demonstrates strong generalization on synthetic data, qualitative tests on real astronomical spectra reveal biases likely stemming from simulator fidelity or training set limitations rather than architectural flaws. This work highlights a critical challenge in scientific machine learning: ensuring model robustness when applied to complex physical systems where ground-truth labels are inherently uncertain.
TNFlow: Amortized Posterior Inference for Trans-Neptunian Object Surface Composition
arXiv:2609.04305v1 Announce Type: cross Abstract: We present TNFlow, a transformer and normalizing flow architecture for inferring the surface composition of Trans-Neptunian Objects (TNOs) from their reflectance spectra. TNFlow is trained on synthetic spectra generated by the Shkuratov radiative transfer model to act as its inverse. TNFlow takes ${\sim}$0.7s to invert one spectrum on a single CPU core, returning a multimodal posterior over simplex-valid compositions and grain sizes. On synthetic spectra, the highest-weight mode achieves a mean total-variation distance of 0.149 from ground truth on the test split, and the model generalizes well to unseen combinations of known components. Qualitative tests on real JWST spectra show blindness or bias towards some materials. We suggest this could be attributed to either simulator fidelity or the training set.
Authors:Agastya Gaur (1 and 2), Cristina M. Dalle Ore (3), Alessandra Ricca (4 and 4) ((1) University of Illinois Urbana-Champaign, (2) SETI Institute, (3) Carl Sagan Center, SETI Institute, (4) NASA Ames Research Center)
Abstract:We present TNFlow, a transformer and normalizing flow architecture for inferring the surface composition of Trans-Neptunian Objects (TNOs) from their reflectance spectra. TNFlow is trained on synthetic spectra generated by the Shkuratov radiative transfer model to act as its inverse. TNFlow takes ${\sim}$0.7s to invert one spectrum on a single CPU core, returning a multimodal posterior over simplex-valid compositions and grain sizes. On synthetic spectra, the highest-weight mode achieves a mean total-variation distance of 0.149 from ground truth on the test split, and the model generalizes well to unseen combinations of known components. Qualitative tests on real JWST spectra show blindness or bias towards some materials. We suggest this could be attributed to either simulator fidelity or the training set.
Instrumentation and Methods for Astrophysics (astro-ph.IM); Earth and Planetary Astrophysics (astro-ph.EP); Machine Learning (cs.LG)
arXiv:2609.04305 [astro-ph.IM]
(or arXiv:2609.04305v1 [astro-ph.IM] for this version)
Text integral sub licența deschisă a publicației · 176 cuvinte
Subiecte conexe
Variational objectives for amortized Bayesian inference in inverse problems: The role of posterior conditioning
arXiv · 23 sept. 2026
Energy-based Transport for Amortized Bayesian Inference
arXiv · 20 iul. 2026
Identifying Informative Environments for Cognition Parameter Inference via Bayesian Experimental Design
arXiv · 3 aug. 2026
Amortizing intractable inference in diffusion models for vision, language, and control
arXiv · 31 aug. 2026