We present a method for measurement analyses based on probabilistic deep neural networks that provide several advantages over conventional analyses with phenomenological models. These include predicting physical quantities directly from data, the rapid generation of statistically robust uncertainties, and the ability to bypass some parameters that may induce ambiguities and complications in data analysis. As deep learning methods make predictions through "black boxes," the uncertainty quantification is typically challenging. We use a probabilistic framework that provides thorough uncertainty quantification and is straightforward to follow in practice. With the network architecture based on the Transformer, we demonstrate the current method for predicting nuclear resonance parameters from scattering data using the phenomenological R-matrix model.

Physics models typically contain adjustable parameters to reproduce measured data. While some parameters correspond directly to measured features in the data, others are unobservable. These unobservables can, in some cases, cause ambiguities in the extraction of observables from measured data, or lead to questions on the physical interpretation of fits that require these extra parameters. We propose a method based on deep learning to extract values of observables directly from the data without the need for unobservables. The key to our approach is to label the training data for the deep learning model with only the observables. After training, the deep learning model can determine the values of observables from measured data with no ambiguities arising from unobservables. We demonstrate this method on the phenomenological R-matrix that is widely utilized in nuclear physics to extract resonance parameters from cross section data. Our deep learning model based on Transformers successfully predicts nuclear properties from measurements with no need for the channel radius and background pole parameters required in traditional R-matrix analyses. Details and limitations of this method, which may be useful for studies of a wide range of phenomena, are discussed.

Physics models typically contain adjustable parameters to reproduce measured data. While some parameters correspond directly to measured features in the data, others are unobservable. These unobservables can, in some cases, cause ambiguities in the extraction of observables from measured data, or lead to questions on the physical interpretation of fits that require these extra parameters. We propose a method based on deep learning to extract values of observables directly from the data without the need for unobservables. The key to our approach is to label the training data for the deep learning model with only the observables. After training, the deep learning model can determine the values of observables from measured data with no ambiguities arising from unobservables. We demonstrate this method on the phenomenological R-matrix that is widely utilized in nuclear physics to extract resonance parameters from cross section data. Our deep learning model based on Transformers successfully predicts nuclear properties from measurements with no need for the channel radius and background pole parameters required in traditional R-matrix analyses. Details and limitations of this method, which may be useful for studies of a wide range of phenomena, are discussed.