Machine Learning Approach in the Prediction of Differential Cross Sections and Structure Functions of Single Pion Electroproduction in the Resonance Regionстатья
Статья опубликована в журнале из списка Web of Science и/или Scopus
Аннотация:This work explores artificial intelligence methods in the task of predicting differential crosssections in exclusive reactions of positively charged pion production induced by virtual photons. A fullyconnected neural network devoid of any prior theoretical knowledge about the scatterring process wastrained on experimental data from the CLAS detector. We present a comparison of the network’s predictionswith experimental data in the form of graphs showing the dependence of differential cross sections onkinematic variables in the excitation energy regions of nucleon resonances, as well as a comparison ofthe structure functions depending on the values of invariant mass of the final hadron system. Based onthis algorithm we can interpolate both the cross-section values and structure function values in differentregions of phase space. The neural network approach preserves all correlations of the multidimensionalspace of kinematic variables, it is model independent and does not consume any a priori knowledge of theprocess, it is easily extensible to a high dimensional space, which can serve as a good basis for buildingMonte Carlo event generators or detailed rection analysis.
