Neutron capture cross-sections are an essential ingredient for many research fields, particularly for nuclear reactor technologies and astrophysical studies. Therefore, a big effort is being made worldwide with the aim of improving neutron-beam facilities, detection systems and sample production techniques with the goal of getting new or more accurate nuclear data.
However, ineluctable experimental effects such as undesired isotope contaminations or misidentification from other nuclear reaction channels remain an important challenge, and many times represent a limiting factor in this type of measurements. This contribution presents a new approach to tackle this challenge based on modern Machine Learning and Artificial Intelligence methods.
Unsupervised Machine Learning techniques implemented using deep neural networks such as Generative Adversarial Neural Networks, Autoencoders and Mutual Information based criterion  can help to disentangle (n,ɣ) signatures of different isotopes registered by the same detection system. They can introduce complementary advantages to the traditional statistical analysis applied to the reaction yield, suppressing undesired backgrounds, identifying isotope nuclear resonances and extracting valuable information about nuclear structure by means of variational learning.
The proposed technique is applicable to any conventional detection system, and it results particularly well performing for new detection systems, such as the total-energy detector with gamma-ray imaging capability, so-called i-TED [2,3], where several variables such as energy, position and time are measured for each registered capture event.
At the time of the conference, a detailed MC study applying these novel techniques will be presented and illustrated aimed at the 94Nb(n,ɣ) experiment that will be carried out in the forthcoming experimental campaign in 2022 at the CERN n_TOF facility .
 Rowel Atienza, Advanced Deep Learning with TensorFlow 2 and Keras (2020) ISBN 978-1-83882-165-4
 V. Babiano-Suarez et al., The European Physical Journal A, Volume 57, Issue 6, article id.197 (2021)
 Project funded by the European Research Council under ERC grant agreement Nr. 681740