arXiv subjects

David Shih

Publications and source records attributed to David Shih.

At least 91 records · Page 5Linked to original sources

arXiv2020

Anomaly Detection with Density Estimation

We leverage recent breakthroughs in neural density estimation to propose a new unsupervised anomaly detection technique (ANODE). By estimating the probability density of the data in a signal region and in sidebands, and interpolating the latter into the signal region, a likelihood ratio of data vs. background can be constructed. This likelihood ratio is broadly sensitive to overdensities in the data that could be due to localized anomalies. In addition, a unique potential benefit of the ANODE method is that the background can be directly estimated using the learned densities. Finally, ANODE is robust against systematic differences between signal region and sidebands, giving it broader applicability than other methods. We demonstrate the power of this new approach using the LHC Olympics 2020 R\&D Dataset. We show how ANODE can enhance the significance of a dijet bump hunt by up to a factor of 7 with a 10\% accuracy on the background prediction. While the LHC is used as the recurring example, the methods developed here have a much broader applicability to anomaly detection in physics and beyond.

hep-ph↗

arXiv2020

Strange Jet Tagging

Tagging jets of strongly interacting particles initiated by energetic strange quarks is one of the few largely unexplored Standard Model object classification problems remaining in high energy collider physics. In this paper we investigate the purest version of this classification problem in the form of distinguishing strange-quark jets from down-quark jets. Our strategy relies on the fact that a strange-quark jet contains on average a higher ratio of neutral kaon energy to neutral pion energy than does a down-quark jet. Long-lived neutral kaons deposit energy mainly in the hadronic calorimeter of a high energy detector, while neutral pions decay promptly to photons that deposit energy mainly in the electromagnetic calorimeter. In addition, short-lived neutral kaons that decay in flight to charged pion pairs can be identified as a secondary vertex in the inner tracking system. Using these handles we study different approaches to distinguishing strange-quark from down-quark jets, including single variable cut-based methods, a boosted decision tree (BDT) with a small number of simple variables, and a deep learning convolutional neural network (CNN) architecture with jet images. We show that modest gains are possible from the CNN compared with the BDT or a single variable. Starting from jet samples with only strange-quark and down-quark jets, the CNN algorithm can improve the strange to down ratio by a factor of roughly 2 for strange tagging efficiencies below 0.2, and by a factor of 2.5 for strange tagging efficiencies near 0.02.

hep-ph↗

arXiv2020

Boosted $W/Z$ Tagging with Jet Charge and Deep Learning

We demonstrate that the classification of boosted, hadronically-decaying weak gauge bosons can be significantly improved over traditional cut-based and BDT-based methods using deep learning and the jet charge variable. We construct binary taggers for $W^+$ vs. $W^-$ and $Z$ vs. $W$ discrimination, as well as an overall ternary classifier for $W^+$/$W^-$/$Z$ discrimination. Besides a simple convolutional neural network (CNN), we also explore a composite of two CNNs, with different numbers of layers in the jet $p_{T}$ and jet charge channels. We find that this novel structure boosts the performance particularly when considering the $Z$ boson as signal. The methods presented here can enhance the physics potential in SM measurements and searches for new physics that are sensitive to the electric charge of weak gauge bosons.

hep-ph↗

arXiv2020

Simulation Assisted Likelihood-free Anomaly Detection

Given the lack of evidence for new particle discoveries at the Large Hadron Collider (LHC), it is critical to broaden the search program. A variety of model-independent searches have been proposed, adding sensitivity to unexpected signals. There are generally two types of such searches: those that rely heavily on simulations and those that are entirely based on (unlabeled) data. This paper introduces a hybrid method that makes the best of both approaches. For potential signals that are resonant in one known feature, this new method first learns a parameterized reweighting function to morph a given simulation to match the data in sidebands. This function is then interpolated into the signal region and then the reweighted background-only simulation can be used for supervised learning as well as for background estimation. The background estimation from the reweighted simulation allows for non-trivial correlations between features used for classification and the resonant feature. A dijet search with jet substructure is used to illustrate the new method. Future applications of Simulation Assisted Likelihood-free Anomaly Detection (SALAD) include a variety of final states and potential combinations with other model-independent approaches.

hep-ph↗

arXiv2019

Searching for muonic forces with the ATLAS detector

The LHC copiously produces muons via different processes, and the muon sample will be large at the high-luminosity LHC (HL-LHC). In this work we propose to leverage this large muon sample and utilize the HL-LHC as a muon fixed-target experiment, with the ATLAS calorimeter as the target. We consider a novel analysis for the ATLAS detector, which takes advantage of the two independent muon momentum measurements by the inner detector and the muon system. We show that a comparison of the two measurements, before and after the calorimeters, can probe new force carriers that are coupled to muons and escape detection. The proposed analysis, based on muon samples from $W$ and $Z$ decays only, has a comparable reach to other proposals. In particular, it can explore the part of parameter-space that could explain the muon $g-2$ anomaly.

hep-ph↗

arXiv2019

Maximizing the Impact of New Physics in $b\rightarrow c τν$ Anomalies

We develop a rigorous, semi-analytical method for maximizing any $b\to cτν$ observable in the full 20-real-dimensional parameter space of the dimension 6 effective Hamiltonian, given some fixed values of $R_{D^{(*)}}$. We apply our method to find the maximum allowed values of $F^L_{D^*}$ and $R_{J/ψ}$, two observables which have both come out higher than their SM predictions in recent measurements by the Belle and LHCb collaborations. While the measurements still have large error bars, they add to the existing $R_{D^{(*)}}$ anomaly, and it is worthwhile to consider NP explanations. It has been shown that none of the existing, minimal models in the literature can explain the observed values of $F^L_{D^*}$ and $R_{J/ψ}$. Using our method, we will generalize beyond the minimal models and show that there is no combination of dimension 6 Wilson operators that can come within $1σ$ of the observed $R_{J/ψ}$ value. By contrast, we will show that the observed value of $F^L_{D^*}$ can be achieved, but only with sizable contributions from tensor and mixed-chirality vector Wilson coefficients.

hep-ph↗

arXiv2019

It's all right(-handed neutrinos): a new $W'$ model for the $R_{D^{(*)}}$ anomaly

The measured $B$-meson semi-leptonic branching ratios $R_{D}$ and $R_{D^*}$ have long-standing deviations between theory and experiment. We introduce a model which explains both anomalies through a single interaction by introducing a right-handed neutrino as the missing energy particle. This interaction is mediated by a heavy charged vector boson ($W'$) which couples only to right-handed quarks and leptons of the Standard Model through the mixing of these particles with new vector-like fermions. Previous $W'$ models for the $R_{D^{(*)}}$ anomaly were strongly constrained from flavor changing neutral currents and direct collider searches for $Z'\toττ$ resonances. We show that relying on right-handed fermion mixing enables us to avoid these constraints, as well as other severe bounds from electroweak precision tests and neutrino mixing.

hep-ph↗

arXiv2019

Asymmetry Observables and the Origin of $R_{D^{(*)}}$ Anomalies

The $R_{D^{(*)}}$ anomalies are among the longest-standing and most statistically significant hints of physics beyond the Standard Model. Many models have been proposed to explain these anomalies, including the interesting possibility that right-handed neutrinos could be involved in the $B$ decays. In this paper, we investigate future measurements at Belle II that can be used to tell apart the various new physics scenarios. Focusing on a number of $τ$ asymmetry observables (forward-backward asymmetry and polarization asymmetries) which can be reconstructed at Belle II, we calculate the contribution of the most general dimension 6 effective Hamiltonian (including right-handed neutrinos) to all of these asymmetries. We show that Belle II can use these asymmetries to distinguish between new-physics scenarios that use right- and left-handed neutrinos, and in most cases can likely distinguish the specific model itself.

hep-ph↗

arXiv2018

Searching for New Physics with Deep Autoencoders

We introduce a potentially powerful new method of searching for new physics at the LHC, using autoencoders and unsupervised deep learning. The key idea of the autoencoder is that it learns to map "normal" events back to themselves, but fails to reconstruct "anomalous" events that it has never encountered before. The reconstruction error can then be used as an anomaly threshold. We demonstrate the effectiveness of this idea using QCD jets as background and boosted top jets and RPV gluino jets as signal. We show that a deep autoencoder can significantly improve signal over background when trained on backgrounds only, or even directly on data which contains a small admixture of signal. Finally we examine the correlation of the autoencoders with jet mass and show how the jet mass distribution can be stable against cuts in reconstruction loss. This may be important for estimating QCD backgrounds from data. As a test case we show how one could plausibly discover 400 GeV RPV gluinos using an autoencoder combined with a bump hunt in jet mass. This opens up the exciting possibility of training directly on actual data to discover new physics with no prior expectations or theory prejudice.

hep-ph↗

arXiv2018

WCxf: an exchange format for Wilson coefficients beyond the Standard Model

We define a data exchange format for numerical values of Wilson coefficients of local operators parameterising low-energy effects of physics beyond the Standard Model. The format facilitates interfacing model-specific Wilson coefficient calculators, renormalisation group (RG) runners, and observable calculators. It is designed to be unambiguous (defining a non-redundant set of operators with fixed normalisation in each basis), extensible (allowing the addition of new EFTs or bases by the user), and robust (being based on industry standard file formats with parsers implemented in many programming languages). We have implemented the format for the Standard Model EFT (SMEFT) and for the weak effective theory (WET) below the electroweak scale and have added interfaces to a number of public codes dealing with SMEFT or WET. We also provide command-line utilities and a Python module for convenient manipulation of WCxf files, including translation between different bases and matching from SMEFT to WET.

hep-ph↗

arXiv2018

Pulling Out All the Tops with Computer Vision and Deep Learning

We apply computer vision with deep learning -- in the form of a convolutional neural network (CNN) -- to build a highly effective boosted top tagger. Previous work (the "DeepTop" tagger of Kasieczka et al) has shown that a CNN-based top tagger can achieve comparable performance to state-of-the-art conventional top taggers based on high-level inputs. Here, we introduce a number of improvements to the DeepTop tagger, including architecture, training, image preprocessing, sample size and color pixels. Our final CNN top tagger outperforms BDTs based on high-level inputs by a factor of $\sim 2$--3 or more in background rejection, over a wide range of tagging efficiencies and fiducial jet selections. As reference points, we achieve a QCD background rejection factor of 500 (60) at 50\% top tagging efficiency for fully-merged (non-merged) top jets with $p_T$ in the 800--900 GeV (350--450 GeV) range. Our CNN can also be straightforwardly extended to the classification of other types of jets, and the lessons learned here may be useful to others designing their own deep NNs for LHC applications.

hep-ph↗

arXiv2017

An Update on the LHC Monojet Excess

In previous work, we identified an anomalous number of events in the LHC jets+MET searches characterized by low jet multiplicity and low-to-moderate transverse energy variables. Here, we update this analysis with results from a new ATLAS search in the monojet channel which also shows a consistent excess. As before, we find that this "monojet excess" is well-described by the resonant production of a heavy colored state decaying to a quark and a massive invisible particle. In the combined ATLAS and CMS data, we now find a local (global) preference of 3.3$σ$ (2.5$σ$) for the new physics model over the Standard Model-only hypothesis. As the signal regions containing the excess are systematics-limited, we consider additional cuts to enhance the signal-to-background ratio. We show that binning finer in $H_T$ and requiring the jets to be more central can increase $S/B$ by a factor of ${\sim} 1.5$.

hep-ph↗

arXiv2017

Cornering Natural SUSY at LHC Run II and Beyond

We derive the latest constraints on various simplified models of natural SUSY with light higgsinos, stops and gluinos, using a detailed and comprehensive reinterpretation of the most recent 13 TeV ATLAS and CMS searches with $\sim 15$ fb$^{-1}$ of data. We discuss the implications of these constraints for fine-tuning of the electroweak scale. While the most "vanilla" version of SUSY (the MSSM with $R$-parity and flavor-degenerate sfermions) with 10% fine-tuning is ruled out by the current constraints, models with decoupled valence squarks or reduced missing energy can still be fully natural. However, in all of these models, the mediation scale must be extremely low ($<100$ TeV). We conclude by considering the prospects for the high-luminosity LHC era, where we expect the current limits on particle masses to improve by up to $\sim 1$ TeV, and discuss further model-building directions for natural SUSY that are motivated by this work.

hep-ph↗

arXiv2017

Digging Deeper for New Physics in the LHC Data

In this paper we describe a novel, model-independent technique of "rectangular aggregations" for mining the LHC data for hints of new physics. A typical (CMS) search now has hundreds of signal regions, which can obscure potentially interesting anomalies. Applying our technique to the two CMS jets+MET SUSY searches, we identify a set of previously overlooked $\sim 3σ$ excesses. Among these, four excesses survive tests of inter- and intra-search compatibility, and two are especially interesting: they are largely overlapping between the jets+MET searches and are characterized by low jet multiplicity, zero $b$-jets, and low MET and $H_T$. We find that resonant color-triplet production decaying to a quark plus an invisible particle provides an excellent fit to these two excesses and all other data -- including the ATLAS jets+MET search, which actually sees a correlated excess. We discuss the additional constraints coming from dijet resonance searches, monojet searches and pair production. Based on these results, we believe the wide-spread view that the LHC data contains no interesting excesses is greatly exaggerated.

hep-ph↗

arXiv2016

Dark Matter and the Higgs in Natural SUSY

Null results from dark matter (DM) direct detection experiments and the 125 GeV Higgs both pose serious challenges to minimal supersymmetry. In this paper, we propose a simple extension of the MSSM that economically solves both problems: a "dark sector" consisting of a singlet and a pair of $SU(2)$ doublets. Loops of the dark sector fields help lift the Higgs mass to 125 GeV consistent with naturalness, while the lightest fermion in the dark sector can be viable thermal relic DM, provided that it is mostly singlet. The DM relic abundance is controlled by s-wave annihilation to tops and Higgsinos, leading to a tight relation between the relic abundance and the spin-dependent direct detection cross section. As a result, the model will be fully probed by the next generation of direct detection experiments. Finally we discuss the discovery potential at LHC Run II.

hep-ph↗

arXiv2016

Precision Corrections to Fine Tuning in SUSY

Requiring that the contributions of supersymmetric particles to the Higgs mass are not highly tuned places upper limits on the masses of superpartners -- in particular the higgsino, stop, and gluino. We revisit the details of the tuning calculation and introduce a number of improvements, including RGE resummation, two-loop effects, a proper treatment of UV vs. IR masses, and threshold corrections. This improved calculation more accurately connects the tuning measure with the physical masses of the superpartners at LHC-accessible energies. After these refinements, the tuning bound on the stop is now also sensitive to the masses of the 1st and 2nd generation squarks, which limits how far these can be decoupled in Effective SUSY scenarios. We find that, for a fixed level of tuning, our bounds can allow for heavier gluinos and stops than previously considered. Despite this, the natural region of supersymmetry is under pressure from the LHC constraints, with high messenger scales particularly disfavored.

hep-ph↗

arXiv2016

FormFlavor Manual

This manual describes the usage and structure of FormFlavor, a Mathematica-based tool for computing a broad list of flavor and CP observables in general new physics models. Based on the powerful machinery of FeynArts and FormCalc, FormFlavor calculates the one-loop Wilson coefficients of the dimension 5 and 6 Standard Model effective Lagrangian entirely from scratch. These Wilson coefficients are then evolved down to the low scale using one-loop QCD RGEs, where they are transformed into flavor and CP observables. The last step is accomplished using a model-independent, largely stand-alone package called FFObservables that is included with FormFlavor. The SM predictions in FFObservables include up-to-date references and accurate current predictions. Using the functions and modular structure provided by FormFlavor, it is straightforward to add new observables. Currently, FormFlavor is set up to perform these calculations for the general, non-MFV MSSM, but in principle it can be generalized to arbitrary FeynArts models. FormFlavor and an up-to-date manual can be downloaded from: http://formflavor.hepforge.org.

hep-ph↗

arXiv2015

General Gauge Mediation at the Weak Scale

We completely characterize General Gauge Mediation (GGM) at the weak scale by solving all IR constraints over the full parameter space. This is made possible through a combination of numerical and analytical methods, based on a set of algebraic relations among the IR soft masses derived from the GGM boundary conditions in the UV. We show how tensions between just a few constraints determine the boundaries of the parameter space: electroweak symmetry breaking (EWSB), the Higgs mass, slepton tachyons, and left-handed stop/sbottom tachyons. While these constraints allow the left-handed squarks to be arbitrarily light, they place strong lower bounds on all of the right-handed squarks. Meanwhile, light EW superpartners are generic throughout much of the parameter space. This is especially the case at lower messenger scales, where a positive threshold correction to $m_h$ coming from light Higgsinos and winos is essential in order to satisfy the Higgs mass constraint.

hep-ph↗