One of the aspects of the understanding of the Universe evolution is its star formation history. In order to gain a complete picture of the Universe evolution it is important to know when the stars we see today were formed. One of the method to study this problem is to use far-infrared and radio emission of galaxies. In this way it is possible to investigate the sites of star formation that are totally obscured by dust and therefore invisible at the optical wavelengths. It is because the energy absorbed by dust in the optical is re-emitted in the infrared, whereas radio emission is unaffected by dust obscuration.

My analysis is based on two samples of galaxies,which have been confirmed to be associated with recent star formation, namely gamma-ray burst (GRB) host galaxies and submillimeterselected galaxies (SMGs). For GRB hosts the long-wavelength data are scarce, so I have started a large observing program targeting these galaxies at the radio wavelengths. The obtained data are analysed simultaneously together with the literature data by means of spectral energy distribution (SED) modelling. I use the radiative transfer code called GRASIL, which calculates the entire UV-to-radio SED of a galaxy taking into account the evolution of stars as well as dust reprocessing in both molecular clouds and diffuse interstellar medium. Using SEDmodelling I explain the seeming discrepancy between long- and short-wavelength properties of the only four GRB hosts that were detected in the submillimeter and/or radio, namely their enhanced submillimeter / radio emission combined with optical faintness and blue colors. I find that these four galaxies are young, highly star-forming, low-mass and dusty.