Modeling complex networks of nuclear reaction data for probing their discovery processes *

Abstract

Hundreds of thousands of experimental data sets of nuclear reactions have been systematically collected, and their number is still growing rapidly. The data and their correlations compose a complex system, which underpins nuclear science and technology. We model the nuclear reaction data as weighted evolving networks for the purpose of data verification and validation. The networks are employed to study the growing cross-section data of a neutron induced threshold reaction (n,2n) and photoneutron reaction. In the networks, the nodes are the historical data, and the weights of the links are the relative deviation between the data points. It is found that the networks exhibit small-world behavior, and their discovery processes are well described by the Heaps law. What makes the networks novel is the mapping relation between the network properties and the salient features of the database: the Heaps exponent corresponds to the exploration efficiency of the specific data set, the distribution of the edge-weights corresponds to the global uncertainty of the data set, and the mean node weight corresponds to the uncertainty of the individual data point. This new perspective to understand the database will be helpful for nuclear data analysis and compilation.

Studies of different kernel functions in nuclear mass predictions with kernel ridge regression

“…Machine learning (ML) has been successfully applied in various fields of physics [24,25]. For nuclear physics, ML applications can be traced back to early 1990s [26,27], and recently, it has been widely adopted to nuclear masses [28][29][30][31][32][33][34][35][36][37][38][39][40][41], charge radii [36,[42][43][44][45], decays and reactions [46][47][48][49][50][51][52][53], ground and excited states [54][55][56][57][58], nuclear landscape [59,60], fission yields [61-63], nuclear liquid-gas phase transition [64], variational calculations [65,66], nuclear energy density functional [67], etc. In nuclear mass studies, ML approaches, such as the radial basis function (RBF) approach [28,29,[68][69][70][71], the Bayesian neural network (BNN) approach [31]…”