Preprint: Log-Gaussian processes for AI-assisted TAS experiments
I am pleased to announce a new preprint containing results of our work on AI-assisted TAS experiments. Together with the PANDA team around Astrid Schneidewind and Christian Franz, Georg Brandl from the MLZ Instrument Control Group, Uwe Stuhr (EIGER instrument responsible at PSI), and Marina Ganeva, leader of the MLZ Data Driven Discovery Group that I am part of, I spent a lot of energy in providing evidence on the benefits and good performance of our approach ARIANE (ARtificial Intelligence-Assisted Neutron Experiments).
The manuscript is an outcome of the project AINX (Artificial Intelligence for Neutron and X-ray scattering) funded by the Helmholtz AI cooperation unit of the German Helmholtz Association.
I would like to thank all the co-authors of this manuscript very much for their help making the good results possible. I am looking forward to and curious about our next steps in the project and how we will approach them.
Abstract. To understand the origins of materials properties, neutron scattering experiments at three-axes spectrometers (TAS) investigate magnetic and lattice excitations in a sample by measuring intensity distributions in its momentum (Q) and energy (E) space. The high demand and limited availability of beam time for TAS experiments however raise the natural question whether we can improve their efficiency or make better use of the experimenter's time. In fact, using TAS, there are a number of scientific questions that require searching for signals of interest in a particular region of Q-E space, but when done manually, it is time consuming and inefficient since the measurement points may be placed in uninformative regions such as the background. Active learning is a promising general machine learning approach that allows to iteratively detect informative regions of signal autonomously, i.e., without human interference, thus avoiding unnecessary measurements and speeding up the experiment. In addition, the autonomous mode allows experimenters to focus on other relevant tasks in the meantime. The approach that we describe in this article exploits log-Gaussian processes which, due to the log transformation, have the largest approximation uncertainties in regions of signal. Maximizing uncertainty as an acquisition function hence directly yields locations for informative measurements. We demonstrate the benefits of our approach on outcomes of a real neutron experiment at the thermal TAS EIGER (PSI) as well as on results of a benchmark in a synthetic setting including numerous different excitations.