Predicting far-infrared images of galaxies

using machine learning techniques

Galaxies emit across the whole wavelength spectrum

In general, more energetic events happen at shorter wavelengths

  • ultraviolet (UV): young, massive stars
  • optical to near-infrared (NIR): older stars
  • mid-to-far-infrared (MIR-FIR): interstellar dust
Andromeda from X-ray to radio

Our neighbor Andromeda, from high-energy X-ray to low-energy radio

My goal is to use the UV-FIR pixels of Andromeda as a training set, in order to learn a mapping from UV-MIR (input) to FIR (output).

Then, even for galaxies where we do not have FIR data, I can predict the FIR image.

I can use the galaxies where FIR is available as a test set.

Find out more in the paper

The project is similar to the one on FIR prediction on global scales.