Statistical Modeling of Nuclear Systematics

Clark, John W.; Mavrommatis, E.; Athanassopoulos, S.; Dakos, A.; Gernoth, K. A.

doi:10.1142/9789812811127_0008

Cited by 4 publications

(5 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our artificial neural networks are trained with a modified version of the SB algorithm [26,28,29] that we have found empirically to be advantageous in the mass-modeling problem. In this new algorithm, to be denoted MB, the weight update prescription corresponding to Eq.…”

Section: Training Algorithmsmentioning

confidence: 99%

“…On the other, quantitative calculation of some properties of some classes of nuclei presents difficult challenges even for the best ab initio quantum-mechanical theories and phenomenological macroscopic/microscopic models. To date, global neuralnetwork models have been developed for the stability/instability dichotomy, for the atomic-mass table, for neutron separation energies, for spins and parities, for decay branching probabilities of nuclear ground states, and for β decay half-lives [20,21,22,23,24,25,26,27,28,29,30].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Nuclear mass systematics using neural networks

et al. 2004

Self Cite

View full text Add to dashboard Cite

New global statistical models of nuclidic (atomic) masses based on multilayered feedforward networks are developed. One goal of such studies is to determine how well the existing data, and only the data, determines the mapping from the proton and neutron numbers to the mass of the nuclear ground state. Another is to provide reliable predictive models that can be used to forecast mass values away from the valley of stability. Our study focuses mainly on the former goal and achieves substantial improvement over previous neural-network models of the mass table by using improved schemes for coding and training. The results suggest that with further development this approach may provide a valuable complement to conventional global models.

show abstract

Section: Training Algorithmsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Nuclear mass systematics using neural networks

et al. 2004

Self Cite

View full text Add to dashboard Cite

show abstract

“…This may be partially ascribed to its larger number of adjustable parameters, although it should be emphasized that in general having more parameters leads to better fitting but not better prediction. Improved performance is also shown by the standard ANN model [1] relative to previous ANN models [17,18]. This improvement is ascribed to its strategic advantages: different architecture and input encoding as well as a more advanced training procedure.…”

Section: Performancementioning

confidence: 90%

“…Statistical global models of this kind have previously been applied to the β-decay problem in Refs. [1,17,18].…”

Section: Introductionmentioning

confidence: 99%

Statistical Global Model of beta- Half-lives and r-Process Nucleosynthesis

Costiris,

Mavrommatis,

Gernoth

et al. 2013

Preprint

Self Cite

View full text Add to dashboard Cite

Background: Reliable prediction of outcomes of stellar nucleosynthesis via the rapid neutron-capture process (r-process) continues to present major challenges to nuclear astrophysics. Uncertainties still persist with respect to the astrophysical sites and the required nuclear-physics inputs. Principally, these inputs involve β-decay rates of neutron-rich nuclei, to which both the element distribution on an r-process path and the time scale of the r-process are highly sensitive. Since the majority of nuclides lying on an r-process path are not yet accessible experimentally, accurate forecasts of nuclear inputs based on well-tested global models are essential.Purpose: Our objective is to apply an improved statistical global model of β − -decay half-life systematics [1] generated by machine-learning techniques to the prediction of β half-lives relevant to r-process nuclei. The primary aim of this application is to complement existing r-process-clock and matter-flow studies, thereby providing additional theoretical support for the planning of future activities of the world's network of rare-isotope laboratories.Method: The statistical model employed in this investigation is rooted in a fully connected, multilayer artificial neural network having feed-forward perceptron architecture [3 − 5 − 5 − 5 − 5 − 1 |116 ]. The network model has been taught the systematics of β − decay based on a subset of the existing data by implementing the optimal Levenberg-Marquardt learning algorithm together with Bayesian regularization and cross-validation. The target domain for the modeling is the half-life systematics of nuclear ground states that decay 100% by the β − mode, with cutoff at 10 6 s.Results: Results are presented for nuclides situated on the r-ladders at N = 50, 82, and 126 where abundances peak, as well as for nuclides that affect abundances between peaks or may be relevant to r-processes under different astrophysical scenarios. The half-lives of some of the targeted neutron-rich nuclides have either been recently measured or will be accessible at rare-isotope laboratories in the relatively near future. The results of our large-scale data-driven half-life calculations (generated by a "theory-thin" global statistical model) are compared to available experimental data, including recent measurements on very neutron-rich nuclei along an r-process path far from the valley of β stability. Comparison is also made with corresponding results from traditional global models derived by semi-phenomenological "theory-thick" approaches.Conclusions: Further evidence is presented that "theory-thin," data-driven statistical global modeling of halflives and other nuclear properties developed by machine-learning techniques can yield reliable predictions that may serve as inputs to different astrophysical scenarios for the r-process.

show abstract

“…In this work, we want to address this challenge and focus on the network robustness and avoidance of multiple solutions. In recent years, artificial neural networks have been used for various extrapolations in nuclear physics [27][28][29][30][31][32][33][34][35], and for the solution of the quantum many-body system [36]. Artificial neural networks use sets of nonlinear functions to describe the complex relationships between input and output variables.…”

Section: Introductionmentioning

confidence: 99%

Extrapolation of nuclear structure observables with artificial neural networks

2019

View full text Add to dashboard Cite

Calculations of nuclei are often carried out in finite model spaces. Thus, finite-size corrections enter, and it is necessary to extrapolate the computed observables to infinite model spaces. In this work, we employ extrapolation methods based on artificial neural networks for observables such as the ground-state energy and the point-proton radius. We extrapolate results from no-core shell model (NCSM) and coupled-cluster (CC) calculations to very large model spaces and estimate uncertainties. Training the network on different data typically yields extrapolation results that cluster around distinct values. We show that a preprocessing of input data, and the inclusion of correlations among the input data, reduces the problem of multiple solutions and yields more stable extrapolated results and consistent uncertainty estimates. We perform extrapolations for groundstate energies and radii in 4 He, 6 Li, and 16 O, and compare the predictions from neural networks with results from infrared extrapolations. arXiv:1905.06317v1 [nucl-th]

show abstract

Statistical Modeling of Nuclear Systematics

Cited by 4 publications

References 14 publications

Nuclear mass systematics using neural networks

Nuclear mass systematics using neural networks

Statistical Global Model of beta- Half-lives and r-Process Nucleosynthesis

Extrapolation of nuclear structure observables with artificial neural networks

Contact Info

Product

Resources

About