"Surya is a 366M-parameter model created by IBM and NASA to forecast solar activity, including flare events, solar winds, and precursors to solar eruptions, which can significantly impact astronaut safety in space as well as terrestrial systems such as communications, power distribution, and more. IBM and NASA trained Surya on nine years of full-resolution (4096x4096 pixel) images from NASA's Solar Dynamic Observatory (SDO) satellite, captured with a 12 minutes cadence."
"Current ML applications in heliophysics research often depends on task-specific data and models trained from scratch, which can be inefficient, prone to overfitting, and limited by the scarcity of labeled data, especially for rare events Ahmadzadeh et al. (2019), which are often the most interesting ones. Despite not being task-specific, Surya outperformed existing specialized models, including U-Net for solar region segmentation, AlexNet for solar flare forecasts, and both AlexNet and ResNet50 for solar wind speed forecasting."
Surya is a 366M-parameter model built to forecast solar activity including flares, solar winds, and eruption precursors that affect astronauts and terrestrial systems. The model was trained on nine years of full-resolution (4096x4096 pixel) Solar Dynamics Observatory imagery captured at 12-minute cadence, enabling learning of fine- and large-scale temporal variability. Surya uses a 2-D transformer architecture with two spectral gating blocks, eight long-short attention blocks, and a decoder for physical-domain reconstruction. Spectral gating combines frequency-domain filtering with learnable weights and adaptive re-weighting to suppress noise and enhance relevant features. Long-short attention captures local dependencies, long-range correlations, and multi-scale representations, and Surya outperformed several task-specific baselines on segmentation, flare forecasting, and solar wind speed prediction.
Read at InfoQ
Unable to calculate read time
Collection
[
|
...
]