Evaluation of Accuracy of Surrogate Models

Abstract

This paper presents a statistical approach to evaluation ofaccuracy ofsurrogate models. A measure ofaccuracy based on a distribution oferrors is introduced. Two issues are discused: validation ofmodel accuracy and verification ofstopping criteria in automatic model generation algorithms exploiting reflective exploration.

The spin structure of the nucleon

“…Experimental verification would be major success for QCD theory. We note that very recently first preliminary data from RHIC on the SSA in di-jet correlations have been reported by the STAR collaboration [85].…”

The Role of Gauge Invariance in Single-Spin Asymmetries

“…First experimental measurements for this process have been performed at RHIC in 2006 [20]. The results seem to indicate that the full QCD treatment, which makes use of gluonic pole cross sections to account for the competing effects of the gluonic initial and final-state interactions, is consistent with the preliminary data, while the 'generalized parton model approach', which only uses partonic cross sections, overestimates the measurements [20].…”

“…Predictions for the quark-Sivers contribution DIS 2007 to these spin-asymmetries in dijet production have been presented [1,18], making use of recently obtained parametrizations [20] for the quark-Sivers first-moments (Qiu-Sterman matrix elements) obtained by fitting to A N data taken at E704, BRAHMS and STAR, and for the quark-Sivers half-moments obtained by fitting to SSA's measured at HERMES. First experimental measurements for this process have been performed at RHIC in 2006 [21]. The results seem to indicate that the full QCD treatment, which makes use of gluonic pole cross sections to account for the competing effects of the gluonic initial and final-state interactions, is consistent with the data, while the 'generalized parton model approach', which only uses partonic cross sections, overestimates the measurements [21].…”

“…First experimental measurements for this process have been performed at RHIC in 2006 [21]. The results seem to indicate that the full QCD treatment, which makes use of gluonic pole cross sections to account for the competing effects of the gluonic initial and final-state interactions, is consistent with the data, while the 'generalized parton model approach', which only uses partonic cross sections, overestimates the measurements [21]. However, the experimental results are at the same time consistent with vanishing SSA's in dijet production and more measurements to increase statistics and disentangle flavor dependence in concurrence with more elaborate theoretical studies would therefore be welcome.…”