June 7, 2024, Friday, 03:00PM (GMT+03.00, Moscow), IKI, Room 200

Abstract:

This study explores the potential of using deep neural networks to detect heterogeneous or atypical structures in astronomical maps. The research was conducted on the cosmic microwave background maps from the Planck mission obtained at various frequencies. The applied approach identified several atypical anomalous structures in the actual maps of the Planck mission. This report provides a detailed description of the machine learning model used and the algorithms for detecting anomalous structures. A map showing the locations of such objects was compiled and compared with the UT78 mask. Model quality assessments were obtained. Examples of model result interpretations are provided, focusing on specific regions of the map. Future research involves expanding the dataset and conducting an in-depth study of the interpretability of model results for a better understanding of the model's decision-making processes.