TCC

This project studies and investigates the use of Deep Learning techniques, such as Convolutional Neural Networks, to detect Galaxy Mergers using astronomical imaging data from photometric surveys. The main questions that arise from this subject are: Is it possible to accurately detect merging galaxies from the photometric data alone? And if so, are we able to get better results by using Deep Learning techniques in comparison with the traditional astronomical methods? And finally, can these principles be applied to detect new Galaxy Mergers in the Southern hemisphere sky, by using data from the S-PLUS project. This work aims to answer those questions and implement solutions for them.

The starting point of this project is an article written by Researchers at ETH Zurich [1] that is called: “Using transfer learning to detect galaxy mergers”. This article uses data collected in the Galaxy Zoo Project [2], that models interacting galaxies and provides astronomical imaging data, and applies it on the detection of merging galaxies by utilizing a deep learning technique called Transfer Learning.

This project requires knowledge of Neural Networks and Deep Learning, which are a subset of Machine Learning. It is important to learn the different concepts surrounding those topics, as well as knowing the tools necessary to properly implement them. A basic understanding of astronomy and the physics of light is required, more specifically galaxy morphology and interactions, and the ideas of electromagnetic waves and wavelength, to be able to discuss idea with astronomers and comprehend the available data.

From this point onwards, the necessary steps needed to complete the project are:

• Understanding the article and testing the results the researchers obtained;
• Studying Deep Learning concepts such as CNN and Transfer Learning;
• Studying Astronomical concepts such as Spectroscopy;
• Studying the FITS data type;
• Using the SDSS data with 3 and 5 channels and applying Deep Learning;
• Using the S-PLUS data and applying Deep Learning;
• Comparing the results;
• Detecting Galaxy Mergers.

Schedule