Ian Whittingham | Searcharxiv

arXiv2019

General classification of light curves using extreme boosting

A significant degree of misclassification of variable stars through the application of machine learning methods to survey data motivates a search for more reliable and accurate machine learning procedures, especially in light of the very large data cubes that will be generated by future surveys and the need for immediate production of accurate, formalised catalogues of variable behaviour to enable science to proceed. In this study, the efficiency of an ensemble machine learning procedure utilising extreme boosting was determined by application to a large sample of data from the OGLE III and IV surveys and from the \textit{Kepler} mission. Through recursive training of classifiers, the study developed a variable star classification workflow which produced an average efficiency determined with the average precision of the model (0.81 for \textit{Kepler} and 0.91 for OGLE) and the $f-score$ of predictions on the test sets. This suggests that extreme boosting can be presented as one of the favourable shallow learning methods in developing a variable star classifier for future large survey projects.

astro-ph.SR↗

arXiv2015

Two-particle irreducible effective actions versus resummation: analytic properties and self-consistency

Approximations based on two-particle irreducible (2PI) effective actions (also known as $Φ$-derivable, Cornwall-Jackiw-Tomboulis or Luttinger-Ward functionals depending on context) have been widely used in condensed matter and non-equilibrium quantum/statistical field theory because this formalism gives a robust, self-consistent, non-perturbative and systematically improvable approach which avoids problems with secular time evolution. The strengths of 2PI approximations are often described in terms of a selective resummation of Feynman diagrams to infinite order. However, the Feynman diagram series is asymptotic and summation is at best a dangerous procedure. Here we show that, at least in the context of a toy model where exact results are available, the true strength of 2PI approximations derives from their self-consistency rather than any resummation. This self-consistency allows truncated 2PI approximations to capture the branch points of physical amplitudes where adjustments of coupling constants can trigger an instability of the vacuum. This, in effect, turns Dyson's argument for the failure of perturbation theory on its head. As a result we find that 2PI approximations perform better than Padé approximation and are competitive with Borel-Padé resummation. Finally, we introduce a hybrid 2PI-Padé method.

hep-th↗