Skip to Main content Skip to Navigation

Stochastic approach to the problem of predictive power in the theoretical modeling of the mean-field

Abstract : Results of our study of the theoretical modelling capacities focussing on the nuclear phenomenological mean-field approaches are presented. It is expected that a realistic theory should be capable of predicting satisfactorily the results of the experiments to come, i.e., having what is called a good predictive power. To study the predictive power of a theoretical model, we had to take into account not only the errors of the experimental data but also the uncertainties originating from approximations of the theoretical formalism and the existence of parametric correlations. One of the central techniques in the parameter adjustment is the solution of what is called the Inverse Problem. Parametric correlations usually induce ill-posedness of the inverse problem; they need to be studied and the model regularised. We have tested two types of realistic phenomenological Hamiltonians showing how to eliminate the parametric correlations theoretically and in practice. We calculate the level confidence intervals, the uncertainty distributions of model predictions and have shown how to improve theory’s prediction capacities and stability.
Document type :
Complete list of metadatas
Contributor : Abes Star :  Contact
Submitted on : Tuesday, March 6, 2018 - 4:20:08 PM
Last modification on : Saturday, June 6, 2020 - 3:24:31 AM
Long-term archiving on: : Thursday, June 7, 2018 - 2:24:51 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01724641, version 1



Irene Dedes Nonell. Stochastic approach to the problem of predictive power in the theoretical modeling of the mean-field. Nuclear Theory [nucl-th]. Université de Strasbourg, 2017. English. ⟨NNT : 2017STRAE017⟩. ⟨tel-01724641⟩



Record views


Files downloads