analiza aiqbench
PhysMetrics.Weather introduces a critical evaluation framework for machine learning weather prediction models that moves beyond standard pixel-based error metrics to assess physical consistency. By quantifying conservation, spectral, and dynamical properties, this tool enables developers to verify if data-driven forecasts adhere to known physical laws before operational deployment. The release addresses a key reliability gap in MLWP, guiding the shift toward physics-informed architectures that balance computational efficiency with scientific validity.
PhysMetrics.Weather: An Evaluation Framework for Physical Consistency in ML Weather Models
arXiv:2606.10642v1 Announce Type: new Abstract: Machine learning weather prediction (MLWP) models have achieved impressive forecasting performance at a small fraction of the computational costs required for traditional physics-based methods. However, they are primarily (1) data-driven and (2) evaluated using pixel-wide error metrics (e.g., RMSE), so there are no guarantees that their forecasts are consistent with known physical laws. We introduce PhysMetrics.Weather, an evaluation framework that assesses the physical realism of MLWP models across three types of metrics: conservation, spectral, and dynamical. By quantifying physical realism, this tool guides the development of physics-informed architectures and helps evaluate whether MLWP models are reliable for operational use. Our framework is available on Github at https://github.com/Emmakast/PhysMetrics.Weather.
Title:PhysMetrics.Weather: An Evaluation Framework for Physical Consistency in ML Weather Models
Abstract:Machine learning weather prediction (MLWP) models have achieved impressive forecasting performance at a small fraction of the computational costs required for traditional physics-based methods. However, they are primarily (1) data-driven and (2) evaluated using pixel-wide error metrics (e.g., RMSE), so there are no guarantees that their forecasts are consistent with known physical laws. We introduce this http URL, an evaluation framework that assesses the physical realism of MLWP models across three types of metrics: conservation, spectral, and dynamical. By quantifying physical realism, this tool guides the development of physics-informed architectures and helps evaluate whether MLWP models are reliable for operational use. Our framework is available on Github at this https URL.
Machine Learning (cs.LG); Atmospheric and Oceanic Physics (physics.ao-ph)
arXiv:2606.10642 [cs.LG]
(or arXiv:2606.10642v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.10642 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 · 221 cuvinte