Towards fully automated micro-IBA data processing using artificial intelligence

[Este seminário será oferecido em inglês]

 

After a short overview of the scientific activities of the LEEL group, the seminar will focuses on an ongoing research project aiming at developing a code suite that should process automatically data acquired from any nuclear microprobe experiment that includes RBS, NRA, ERDA, PIXE or any simultaneous association of them. The objective is to reduce significantly the time required by IBA scientists to get reliable results from their experiments. The classical data reduction processing imposes by itself numerous steps, but more importantly, requires multiple and time-consuming iterations between fits of the spectra issued from the simultaneously collected techniques, until a satisfactory coherence is reached. During the seminar, we will present the used approach based on statistical methods and machine learning algorithms. We will particularly emphasize on data augmentation and optimized neural network topology, two techniques that we found able to enhance considerably the global execution time.

 

Sobre o palestrante: O pesquisador H. Khodja é atualmente diretor do Laboratório de Estudos de Elementos Leves na Universidade Paris-Saclay. Seus principais interesses de pesquisa residem em Física de feixe de íons; Comportamento de elementos leves em materiais; Física e biologia da radiação. Acesse AQUI para mais informações sobre o pesquisador.