École Nationale Supérieure de
Techniques Avancées

Abstract

The work presented in the following pages is focused on using machine learning algorithms to shape a custom spectrum resulting from the filamentation of a laser beam in an ethanol cuvette. The algorithm of choice is a densely connected feed forward neural network. The idea is to select a fixed number of empirical parameters that describe the essence of the laser beam used, those parameters are then put in correlation with the shapes of the recorded spectra thanks to the machine learning program. The elaboration of a dataset of the right configuration and the implementation of the neural network are the main focuses discussed in this traineeship record. What is not showcased in this record is the development of a web-based version of the code to enables more flexibility for the users. The approach pursed is to give step-by-step insights that lead to a first achievement in this field of research.