The DNNLikelihood: enhancing likelihood distribution with Deep Learning
Authors:
Andrea Coccaro,
Maurizio Pierini,
Luca Silvestrini,
Riccardo Torre
Abstract:
We introduce the DNNLikelihood, a novel framework to easily encode, through Deep Neural Networks (DNN), the full experimental information contained in complicated likelihood functions (LFs). We show how to efficiently parametrise the LF, treated as a multivariate function of parameters and nuisance parameters with high dimensionality, as an interpolating function in the form of a DNN predictor. We…
▽ More
We introduce the DNNLikelihood, a novel framework to easily encode, through Deep Neural Networks (DNN), the full experimental information contained in complicated likelihood functions (LFs). We show how to efficiently parametrise the LF, treated as a multivariate function of parameters and nuisance parameters with high dimensionality, as an interpolating function in the form of a DNN predictor. We do not use any Gaussian approximation or dimensionality reduction, such as marginalisation or profiling over nuisance parameters, so that the full experimental information is retained. The procedure applies to both binned and unbinned LFs, and allows for an efficient distribution to multiple software platforms, e.g. through the framework-independent ONNX model format. The distributed DNNLikelihood can be used for different use cases, such as re-sampling through Markov Chain Monte Carlo techniques, possibly with custom priors, combination with other LFs, when the correlations among parameters are known, and re-interpretation within different statistical approaches, i.e. Bayesian vs frequentist. We discuss the accuracy of our proposal and its relations with other approximation techniques and likelihood distribution frameworks. As an example, we apply our procedure to a pseudo-experiment corresponding to a realistic LHC search for new physics already considered in the literature.
△ Less
Submitted 5 December, 2019; v1 submitted 8 November, 2019;
originally announced November 2019.
On Flavourful Easter eggs for New Physics hunger and Lepton Flavour Universality violation
Authors:
Marco Ciuchini,
Antonio M. Coutinho,
Marco Fedele,
Enrico Franco,
Ayan Paul,
Luca Silvestrini,
Mauro Valli
Abstract:
Within the standard approach of effective field theory of weak interactions for $ΔB = 1$ transitions, we look for possibly unexpected subtle New Physics effects, here dubbed "flavourful Easter eggs". We perform a Bayesian global fit using the publicly available HEPfit package, taking into account state-of-the-art experimental information concerning these processes, including the suggestive measure…
▽ More
Within the standard approach of effective field theory of weak interactions for $ΔB = 1$ transitions, we look for possibly unexpected subtle New Physics effects, here dubbed "flavourful Easter eggs". We perform a Bayesian global fit using the publicly available HEPfit package, taking into account state-of-the-art experimental information concerning these processes, including the suggestive measurements from LHCb of $R_{K}$ and $R_{K^{*}}$, the latter available only very recently. We parametrize New Physics contributions to $b \to s$ transitions in terms of shifts of Wilson coefficients of the electromagnetic dipole and semi-leptonic operators, assuming CP-conserving effects, but allowing in general for violation of lepton flavour universality. We show how optimistic/conservative hadronic estimates can impact quantitatively the size of New Physics extracted from the fit. With a conservative approach to hadronic uncertainties we find nonzero New Physics contributions to Wilson coefficients at the level of $\sim 3σ$, depending on the model chosen. Furthermore, given the interplay between hadronic contributions and New Physics effects in the leptonic vector current, a scenario with nonstandard leptonic axial currents is comparable to the more widely advocated one with New Physics in the leptonic vector current.
△ Less
Submitted 3 October, 2017; v1 submitted 18 April, 2017;
originally announced April 2017.
