analiza aiqbench
SCORE introduces a scalable framework for approximating covariance matrices in high-dimensional Gaussian models, addressing efficiency and stability through spectral decomposition. By decoupling marginal distributions from correlation structures, the method achieves linear storage and reduced computational costs while maintaining numerical robustness even with degenerate covariances. This approach offers significant practical benefits for complex predictive modeling tasks like time-series forecasting and weather prediction where Gaussian assumptions are standard.
SCORE: Spectral Correlation Estimation for Multivariate Gaussians
arXiv:2610.12096v1 Announce Type: new Abstract: Neural network-based predictive modeling with high-dimensional structured Gaussian targets requires an efficient and numerically stable, yet expressive approximation of the covariance matrix. We propose SCORE: a scalable framework, combining scoring rule training with an expressive covariance approximation learned in spectral space. For $d$-dimensional data, the learning task is decomposed into learning the marginal distributions and learning a structured correlation matrix, which enables dense dependencies with linear storage and $\mathcal{O}(d\log d)$ cost. We utilize the closed form Gaussian kernel score for training, which remains defined even for degenerate covariances and admits bounded gradients during optimization. We characterize kernel scores under invertible transforms and prove exact invariance under unitary transforms. At population level, our two-level objective recovers the true marginals and projects the target correlation onto the representable class; finite-sample PAC bounds show that the errors of the two stages enter additively. We evaluate our model on a variety of tasks with a commonly assumed Gaussian domain: Time-series forecasting, monocular depth estimation, and spatial weather prediction, showing improved performance at lower computational cost.
Abstract:Neural network-based predictive modeling with high-dimensional structured Gaussian targets requires an efficient and numerically stable, yet expressive approximation of the covariance matrix. We propose SCORE: a scalable framework, combining scoring rule training with an expressive covariance approximation learned in spectral space. For $d$-dimensional data, the learning task is decomposed into learning the marginal distributions and learning a structured correlation matrix, which enables dense dependencies with linear storage and $\mathcal{O}(d\log d)$ cost. We utilize the closed form Gaussian kernel score for training, which remains defined even for degenerate covariances and admits bounded gradients during optimization. We characterize kernel scores under invertible transforms and prove exact invariance under unitary transforms. At population level, our two-level objective recovers the true marginals and projects the target correlation onto the representable class; finite-sample PAC bounds show that the errors of the two stages enter additively. We evaluate our model on a variety of tasks with a commonly assumed Gaussian domain: Time-series forecasting, monocular depth estimation, and spatial weather prediction, showing improved performance at lower computational cost.
Machine Learning (cs.LG)
arXiv:2610.12096 [cs.LG]
(or arXiv:2610.12096v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.12096 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 265 cuvinte
Subiecte conexe
Quantization Error Is Spectrally Flat: A Single Random Probe Is a Calibrated, Data-Free Sensitivity Estimator, with Application to Budget-Targeted Mixed-Precision Quantization
arXiv · 29 sept. 2026
Differentially Private Nonparametric Modal Learning with Applications to Regression and Clustering
arXiv · 3 aug. 2026
A Mike's-Eye View of ARC's Research
AI Alignment Forum · 9 iun. 2026
Recursively Trained Diffusion Models: Limiting Collapse Distribution and Spectral Characterization
arXiv · 15 iun. 2026