KASI AI Hub

CAAC Website Launched

The Center for AI & Advanced Computing now has a dedicated website for sharing papers, events, news, and artifacts with institute members.

A new deep-learning approach to infer solar and geomagnetic parameters for the 1859 Carrington event

In this study, we investigate the solar and geomagnetic parameters of the 1859 Carrington event using deep learning and empirical relationships. For this, we apply an image translation model, a popular deep learning method based on conditional Generative Adversarial Networks, to the generation of magnetograms from sunspot drawings. We train the model using pairs of sunspot data from Debrecen Photoheliographic Data and their corresponding Solar and Heliospheric Observatory/Michelson Doppler Imager (SOHO/MDI) and Solar Dynamics Observatory/Helioseismic and Magnetic Imager (SDO/HMI) magnetograms from 1996 to 2018, using data from January-July and December of each year for training and data from August and November for validation. To test the model, we compare actual magnetograms with artificial-intelligence-based (AI-based) ones for September and October. Our results show that the unsigned magnetic fluxes of AI-based magnetograms closely match those of the originals. Applying this model to Carrington’s full-disk sunspot drawing of 1 September 1859, we generate an AI-based magnetogram and estimate its unsigned magnetic flux. To estimate solar and geomagnetic parameters, we use the following empirical relationships: magnetic flux and flare peak flux, magnetic flux and coronal mass ejection (CME) speed, CME speed and transit time, CME speed and interplanetary coronal mass ejection (ICME) speed, and ICME speed and the Disturbance Storm Time (Dst) index to obtain upper-limit estimates for an extreme event. We find that the estimated Sun-Earth transit time is 16.7 h, consistent with the historical observations. The corresponding Dst value is about -1313 nT, which is broadly consistent with previous reconstruction-based estimates for the Carrington storm.

A Convolutional Neural Network-Transformer Denoiser for Low-signal-to-noise-ratio Galaxy Spectra: Stellar Population Recovery in Synthetic Tests

Stellar population measurements in integral field unit surveys are often limited by low signal-to-noise ratios(S/Ns) in low-surface-brightness spaxels. Using controlled synthetic experiments, we investigate whether a deeplearning-based denoising can recover stellar population information from such spectra without requiring spatial binning. We introduce the Enhanced U-Net Transformer (EUT), a one-dimensional convolutional neural network-transformer model trained on 90,000 synthetic spectra constructed from MILES simple stellar population (SSP) models following J. H. Lee et al., with wavelength-dependent noise injected on the fly to emulate SAMI-like data (S/N - 5-20, measured in a 4484.77-4573.12 A continuum window). Utilizing an independent test set of 10,000 spectra, the EUT reduces the full-spectrum rms residual by -96.5% at S/N = 5(and by -94% at S/N = 20), achieving recovery rates of -99.8% (the Pearson correlation coefficient between the noise-free and comparison spectra expressed in percent). In fixed windows around Ca II H, Hδ, Hβ, Fe I 4383, Mg b, and Na D, residuals decrease by -88% while preserving line-profile structure. In downstream analysis withPPXF we assess parameter recovery using the Pearson correlation coefficient Rp and the rms scatter: the scatter in recovered mass-weighted age decreases from -0.41 to -0.25 dex at S/N = 5 and from -0.32 to -0.22 dex at S/N = 10; the corresponding mass-weighted global metallicity, [M/H], scatter decreases from -0.45 to-0.36 dex and from -0.32 to -0.28 dex. At S/N = 20, denoising yields results consistent with those from the noisy inputs within the synthetic-test uncertainties. These controlled experiments suggest that hybrid CNN-transformer denoisers can enhance the usable low-surface-brightness area for stellar population studies, although further validation with observed spectra will be needed before practical application.

The SPHEREx Ices Investigation: An Overview

SPHEREx is a NASA mission designed to perform an all-sky spectroscopic survey in the 0.75-5 μm wavelength range. Its primary science objectives are to investigate: (1) inflationary cosmology; (2) the history of galaxy formation; and (3) the abundance of molecular ices-critical for prebiotic chemistry-found on the surfaces of interstellar dust grains within planet-forming regions. This paper focuses on the third theme, the SPHEREx Ices Investigation, for which SPHEREx is conducting a spectroscopic survey of nearly 10 million preselected sources throughout the Milky Way and Magellanic Clouds, to characterize their ice absorption features. By selecting targets based on infrared color, spatial isolation, and brightness, the Ices Investigation secures high-signal-to-noise-ratio spectra across a broad range of astrophysical environments that are relatively free of spectral contamination. Rather than attempting to decompose each spectrum into its individual ice components, the Ices Investigation prioritizes accurate measurements of the integrated optical depths of key molecular ice absorption features. This approach enables statistically powerful correlation studies between ice abundances and environmental parameters-including extinction, temperature, gas composition, radiation field strength, cosmic-ray flux, and star formation activity. The data pipeline developed for this purpose incorporates machine learning for continuum estimation, drawing on both SPHEREx and ancillary data sets. Ultimately, the expansive spectral archive produced by SPHEREx, combined with targeted follow-up from facilities like JWST, will transform our understanding of Galactic ice formation, evolution, abundance, and their inheritance into planetary systems and prebiotic inventories.