Can artificial intelligence techniques replace human labor in EXAFS data analysis?

Extended X-ray absorption fine structure (EXAFS) is a unique tool used to describe local structural properties around a selected element in a material. However, the quantitative analysis of EXAFS data remains a non-trivial task, especially for beginners in the field of EXAFS. While AI techniques can assist in the analysis of EXAFS data, there are still numerous challenges to overcome for the complete automation of EXAFS data analysis. We explored the automatic analysis of EXAFS data from various materials using deep reinforcement learning (DRL) methods. Unlike other AI techniques, DRL methods do not necessitate a large amount of pre-prepared data to train the neural networks (NNs) of an AI system, as they achieve optimal fits of EXAFS data by using a reward value set as the reciprocal of the R-factor of an EXAFS data fit. Our results strongly indicate that the DRL-EXAFS method can quantitatively analyze EXAFS data.