Speaker
Description
Research in Diffuse Interstellar Bands (DIBs) has been ongoing for over a hundred years, but still little is known about their carrier molecules. Now a new study tries to shed a light on this. The EDIBLES study collects high quality spectra of DIBs in over a hundred different sightlines. Since the inception of the study in 2017 over 10 papers have been published using the EDIBLES study as a base.
The idea of the EDIBLES study is to collect more data on DIBs and to explore new ways of studying them. Some of the results that have already been found are: The identification of several DIB profile families indicating chemically similar carriers and the confirmation of C60+ as a carrier of DIBs. A new project tries to identify DIBs in a spectrum using machine learning techniques.
The machine learning technique used is a random forest. A random forest relies on multiple decision trees to make a prediction of a specific DIB-profile. We will compare this prediction with the prediction of a skewed Gaussian as described in EDIBLES survey VIII. This could determine whether random forest is a viable and faster method for identifying DIBs compared to using Gaussian-fits.
| Talk category | data science (contact d.huppenkothen@uva.nl) |
|---|---|
| Preference for a talk or poster | Poster |
