2026-04-14T07:42:35Zhttps://recercat.cat/oai/request

oai:recercat.cat:2117/4439902026-02-07T23:22:15Zcom_2072_1033col_2072_452951

Feasibility analysis of neural network architectures for signal processing in space missions Zein, Topias Ali Universitat Politècnica de Catalunya. Física Torres Gil, Santiago García Zamora, Enrique Miguel Àrees temàtiques de la UPC::Aeronàutica i espai Artificial Intelligence Neural networks (Computer science) Graphics processing units Artificial Neural Space Astrophysics Networks Intelligence Mission Intel·ligència artificial Xarxes neuronals (Informàtica) Processadors gràfics Artificial Intelligence has become an essential tool in modern science and technology. In particular, deep learning, based on neural network architectures, is one of the most effective and widely used strategies. Simultaneously, current space missions, particularly those focused on astronomy and astrophysics, generate an overwhelming quantity of data that exceeds human capacity for direct analysis and inspection. In this project, we aim to analyse the feasibility of neural network architectures in the context of space missions. A variety of archetypal signals, such as Gaussian distributions and blackbody spectra, will be studied, both in the absence and presence of different types of noise. The analysis will focus on the impact of the number of neurons and layers on the networks' performance, as well as the computational efficiency in terms of processing time using GPU-based systems. 9 - Indústria, Innovació i Infraestructura 2025-10-13 Master thesis https://hdl.handle.net/2117/443990 PRISMA-197739 eng Open Access application/pdf Universitat Politècnica de Catalunya