Neural Network Predictions
Based on supervised deep learning algorithms, the CADE service produced dust emission estimates at 850 microns (353 GHz) and 1.38 mm (217 GHz) at the Herschel resolution (37 arcsec). To perform this work, Herschel data of Galactic environments, smoothed to the 5 arcmin angular resolution, have been used to train the neural network to provide the best model able to reproduce Planck maps of dust emission (after removing CMB, CIB and 12CO contributions) in the two bands centered at the two previous wavelengths. Then, using Herschel data only, the model is applied to predict dust emission maps at 37 arcsec.
The service provides to the community dust emission maps of several surveys: Hi-GAL, Gould Belt, Cold Cores, HERITAGE, KINGFISH, Helga, HerM33es and VNGS. Planck maps after CMB, CIB and 12CO contributions removal are also available at the end of this page.
The units of the Healpix maps are MJy/sr.
References
- 'Inferring the Dust Emission in the sub-millimeter and millimeter wavelengths using Neural Networks' by Paradis, D. et al, 2024, A&A, accepted
Prediction Maps at 37 arcsec and 5 arcmin
Planck Maps at 5 arcmin (after removing CMB, CIB and 12 CO contributions)
850 microns (353 GHz) : planck_850_allsub.fits
1.38 mm (353 GHz) : planck_1p4_allsub.fits