arXiv subjects

Johann Brehmer

Publications and source records attributed to Johann Brehmer.

35 records · Page 2Linked to original sources

arXiv2020

The frontier of simulation-based inference

Many domains of science have developed complex simulations to describe phenomena of interest. While these simulations provide high-fidelity models, they are poorly suited for inference and lead to challenging inverse problems. We review the rapidly developing field of simulation-based inference and identify the forces giving new momentum to the field. Finally, we describe how the frontier is expanding so that a broad audience can appreciate the profound change these developments may have on science.

stat.ML↗

arXiv2020

MadMiner: Machine learning-based inference for particle physics

Precision measurements at the LHC often require analyzing high-dimensional event data for subtle kinematic signatures, which is challenging for established analysis methods. Recently, a powerful family of multivariate inference techniques that leverage both matrix element information and machine learning has been developed. This approach neither requires the reduction of high-dimensional data to summary statistics nor any simplifications to the underlying physics or detector response. In this paper we introduce MadMiner, a Python module that streamlines the steps involved in this procedure. Wrapping around MadGraph5_aMC and Pythia 8, it supports almost any physics process and model. To aid phenomenological studies, the tool also wraps around Delphes 3, though it is extendable to a full Geant4-based detector simulation. We demonstrate the use of MadMiner in an example analysis of dimension-six operators in ttH production, finding that the new techniques substantially increase the sensitivity to new physics.

hep-ph↗

arXiv2019

Benchmarking simplified template cross sections in $WH$ production

Simplified template cross sections define a framework for the measurement and dissemination of kinematic information in Higgs measurements. We benchmark the currently proposed setup in an analysis of dimension-6 effective field theory operators for $WH$ production. Calculating the Fisher information allows us to quantify the sensitivity of this framework to new physics and study its dependence on phase space. New machine-learning techniques let us compare the simplified template cross section framework to the full, high-dimensional kinematic information. We show that the way in which we truncate the effective theory has a sizable impact on the definition of the optimal simplified template cross sections.

hep-ph↗

arXiv2019

Mining for Dark Matter Substructure: Inferring subhalo population properties from strong lenses with machine learning

The subtle and unique imprint of dark matter substructure on extended arcs in strong lensing systems contains a wealth of information about the properties and distribution of dark matter on small scales and, consequently, about the underlying particle physics. However, teasing out this effect poses a significant challenge since the likelihood function for realistic simulations of population-level parameters is intractable. We apply recently-developed simulation-based inference techniques to the problem of substructure inference in galaxy-galaxy strong lenses. By leveraging additional information extracted from the simulator, neural networks are efficiently trained to estimate likelihood ratios associated with population-level parameters characterizing substructure. Through proof-of-principle application to simulated data, we show that these methods can provide an efficient and principled way to simultaneously analyze an ensemble of strong lenses, and can be used to mine the large sample of lensing images deliverable by near-future surveys for signatures of dark matter substructure.

astro-ph.CO↗

arXiv2019

Mining gold from implicit models to improve likelihood-free inference

Simulators often provide the best description of real-world phenomena. However, they also lead to challenging inverse problems because the density they implicitly define is often intractable. We present a new suite of simulation-based inference techniques that go beyond the traditional Approximate Bayesian Computation approach, which struggles in a high-dimensional setting, and extend methods that use surrogate models based on neural networks. We show that additional information, such as the joint likelihood ratio and the joint score, can often be extracted from simulators and used to augment the training data for these surrogate models. Finally, we demonstrate that these new techniques are more sample efficient and provide higher-fidelity inference than traditional methods.

stat.ML↗

arXiv2019

Effective LHC measurements with matrix elements and machine learning

One major challenge for the legacy measurements at the LHC is that the likelihood function is not tractable when the collected data is high-dimensional and the detector response has to be modeled. We review how different analysis strategies solve this issue, including the traditional histogram approach used in most particle physics analyses, the Matrix Element Method, Optimal Observables, and modern techniques based on neural density estimation. We then discuss powerful new inference methods that use a combination of matrix element information and machine learning to accurately estimate the likelihood function. The MadMiner package automates all necessary data-processing steps. In first studies we find that these new techniques have the potential to substantially improve the sensitivity of the LHC legacy measurements.

hep-ph↗

arXiv2018

Likelihood-free inference with an improved cross-entropy estimator

We extend recent work (Brehmer, et. al., 2018) that use neural networks as surrogate models for likelihood-free inference. As in the previous work, we exploit the fact that the joint likelihood ratio and joint score, conditioned on both observed and latent variables, can often be extracted from an implicit generative model or simulator to augment the training data for these surrogate models. We show how this augmented training data can be used to provide a new cross-entropy estimator, which provides improved sample efficiency compared to previous loss functions exploiting this augmented training data.

stat.ML↗

arXiv2018

A Guide to Constraining Effective Field Theories with Machine Learning

We develop, discuss, and compare several inference techniques to constrain theory parameters in collider experiments. By harnessing the latent-space structure of particle physics processes, we extract extra information from the simulator. This augmented data can be used to train neural networks that precisely estimate the likelihood ratio. The new methods scale well to many observables and high-dimensional parameter spaces, do not require any approximations of the parton shower and detector response, and can be evaluated in microseconds. Using weak-boson-fusion Higgs production as an example process, we compare the performance of several techniques. The best results are found for likelihood ratio estimators trained with extra information about the score, the gradient of the log likelihood function with respect to the theory parameters. The score also provides sufficient statistics that contain all the information needed for inference in the neighborhood of the Standard Model. These methods enable us to put significantly stronger bounds on effective dimension-six operators than the traditional approach based on histograms. They also outperform generic machine learning methods that do not make use of the particle physics structure, demonstrating their potential to substantially improve the new physics reach of the LHC legacy results.

hep-ph↗

arXiv2018

Constraining Effective Field Theories with Machine Learning

We present powerful new analysis techniques to constrain effective field theories at the LHC. By leveraging the structure of particle physics processes, we extract extra information from Monte-Carlo simulations, which can be used to train neural network models that estimate the likelihood ratio. These methods scale well to processes with many observables and theory parameters, do not require any approximations of the parton shower or detector response, and can be evaluated in microseconds. We show that they allow us to put significantly stronger bounds on dimension-six operators than existing methods, demonstrating their potential to improve the precision of the LHC legacy constraints.

hep-ph↗

arXiv2017

Better Higgs-CP Tests Through Information Geometry

Measuring the CP symmetry in the Higgs sector is one of the key tasks of the LHC and a crucial ingredient for precision studies, for example in the language of effective Lagrangians. We systematically analyze which LHC signatures offer dedicated CP measurements in the Higgs-gauge sector, and discuss the nature of the information they provide. Based on the Fisher information measure, we compare the maximal reach for CP-violating effects in weak boson fusion, associated ZH production, and Higgs decays into four leptons. We find a subtle balance between more theory-independent approaches and more powerful analysis channels, indicating that rigorous evidence for CP violation in the Higgs-gauge sector will likely require a multi-step process.

hep-ph↗

arXiv2017

Better Higgs Measurements Through Information Geometry

Information geometry can be used to understand and optimize Higgs measurements at the LHC. The Fisher information encodes the maximum sensitivity of observables to model parameters for a given experiment. Applied to higher-dimensional operators, it defines the new physics reach of any LHC signature. We calculate the Fisher information for Higgs production in weak boson fusion with decays into tau pairs and four leptons, and for Higgs production in association with a single top quark. In a next step we analyze how the differential information is distributed over phase space, which defines optimal event selections. Conversely, we consider the information in the distribution of a subset of the kinematic variables, showing which production and decay observables are the most powerful and how much information is lost in traditional histogram-based analysis methods compared to fully multivariate ones.

hep-ph↗

arXiv2016

Pushing Higgs Effective Theory over the Edge

Based on a vector triplet model we study a possible failure of dimension-6 operators in describing LHC Higgs kinematics. First, we illustrate that including dimension-6 contributions squared can significantly improve the agreement between the full model and the dimension-6 approximation, both in associated Higgs production and in weak-boson-fusion Higgs production. Second, we test how a simplified model with an additional heavy scalar could improve the agreement in critical LHC observables. In weak boson fusion we find an improvement for virtuality-related observables at large energies, but at the cost of sizeable deviations in interference patterns and angular correlations.

hep-ph↗

arXiv2016

Pushing Higgs Effective Theory to its Limits

At the LHC, an effective theory of the Higgs sector allows us to analyze kinematic distributions in addition to inclusive rates, although there is no clear hierarchy of scales. We systematically analyze how well dimension-6 operators describe LHC observables in comparison to the full theory, and in a range where the LHC will be sensitive. The key question is how the breakdown of the dimension-6 description affects Higgs measurements during the upcoming LHC run for weakly interacting models. We cover modified Higgs sectors with a singlet and doublet extension, new top partners, and a vector triplet. First, weakly interacting models only generate small relevant subsets of dimension-6 operators. Second, the dimension-6 description tends to be justified at the LHC. Scanning over model parameters, significant discrepancies can nevertheless arise; their main source is the matching procedure in the absence of a well-defined hierarchy of scales. This purely theoretical problem should not affect future LHC analyses.

hep-ph↗

arXiv2015

The Diboson Excess: Experimental Situation and Classification of Explanations; A Les Houches Pre-Proceeding

We examine the `diboson' excess at $\sim 2$ TeV seen by the LHC experiments in various channels. We provide a comparison of the excess significances as a function of the mass of the tentative resonance and give the signal cross sections needed to explain the excesses. We also present a survey of available theoretical explanations of the resonance, classified in three main approaches. Beyond that, we discuss methods to verify the anomaly, determining the major properties of the various surpluses and exploring how different models can be discriminated. Finally, we give a tabular summary of the numerous explanations, presenting their main phenomenological features.

hep-ph↗

arXiv2015

Symmetry Restored in Dibosons at the LHC?

A number of LHC resonance search channels display an excess in the invariant mass region of 1.8 - 2.0 TeV. Among them is a $3.4\,σ$ excess in the fully hadronic decay of a pair of Standard Model electroweak gauge bosons, in addition to potential signals in the $HW$ and dijet final states. We perform a model-independent cross-section fit to the results of all ATLAS and CMS searches sensitive to these final states. We then interpret these results in the context of the Left-Right Symmetric Model, based on the extended gauge group $SU(2)_L\times SU(2)_R\times U(1)'$, and show that a heavy right-handed gauge boson $W_R$ can naturally explain the current measurements with just a single coupling $g_R \sim 0.4$. In addition, we discuss a possible connection to dark matter.

hep-ph↗

arXiv2014

Polarized WW Scattering on the Higgs Pole

The Higgs discovery has given us the Higgs-gauge sector as a new handle to search for physics beyond the Standard Model. This includes physics scenarios originally linked to massive gauge boson scattering at high energies. We investigate how one can separately probe the Higgs couplings to the longitudinal and transverse parts of the massive gauge bosons away from this high-energy limit. Deviations from the Standard Model could originate from higher-dimensional operators, compositeness, or even more fundamentally from a violation of gauge invariance. The signature we propose is the tagging jet kinematics in weak boson scattering for scattering energies close to the Higgs resonance. During the upcoming LHC run at 13 TeV we will be able to test these couplings at the 20% level.

hep-ph↗

arXiv2012

Monte-Carlo event generation for a two-Higgs-doublet model with maximal CP symmetry

Recently a two-Higgs-doublet model with maximal symmetry under generalised CP transformations, the MCPM, has been proposed. The theory features a unique fermion mass spectrum which, although not describing nature precisely, provides a good approximation. It also predicts the existence of five Higgs bosons with a particular signature. In this thesis I implemented the MCPM into the Monte-Carlo event generation package MadGraph, allowing the simulation of any MCPM tree-level process. The generated events are in a standardised format and can be used for further analysis with tools such as PYTHIA or GEANT, eventually leading to the comparison with experimental data and the exclusion or discovery of the theory. The implementation was successfully validated in different ways. It was then used for a first comparison of the MCPM signal events with the SM background and previous searches for new physics, hinting that the data expected at the LHC in the next years might provide exclusion limits or show signatures of this model.

hep-ph↗