arXiv subjects

Leandro G. Almeida

Publications and source records attributed to Leandro G. Almeida.

13 recordsLinked to original sources

arXiv2019

Regression Constraint for an Explainable Cervical Cancer Classifier

This article adresses the problem of automatic squamous cells classification for cervical cancer screening using Deep Learning methods. We study different architectures on a public dataset called Herlev dataset, which consists in classifying cells, obtained by cervical pap smear, regarding the severity of the abnormalities they represent. Furthermore, we use an attribution method to understand which cytomorphological features are actually learned as discriminative to classify severity of the abnormalities. Through this paper, we show how we trained a performant classifier: 74.5\% accuracy on severity classification and 94\% accuracy on normal/abnormal classification.

eess.IV↗

arXiv2015

Playing Tag with ANN: Boosted Top Identification with Pattern Recognition

Many searches for physics beyond the Standard Model at the Large Hadron Collider (LHC) rely on top tagging algorithms, which discriminate between boosted hadronic top quarks and the much more common jets initiated by light quarks and gluons. We note that the hadronic calorimeter (HCAL) effectively takes a "digital image" of each jet, with pixel intensities given by energy deposits in individual HCAL cells. Viewed in this way, top tagging becomes a canonical pattern recognition problem. With this motivation, we present a novel top tagging algorithm based on an Artificial Neural Network (ANN), one of the most popular approaches to pattern recognition. The ANN is trained on a large sample of boosted tops and light quark/gluon jets, and is then applied to independent test samples. The ANN tagger demonstrated excellent performance in a Monte Carlo study: for example, for jets with p_T in the 1100-1200 GeV range, 60% top-tag efficiency can be achieved with a 4% mis-tag rate. We discuss the physical features of the jets identified by the ANN tagger as the most important for classification, as well as correlations between the ANN tagger and some of the familiar top-tagging observables and algorithms.

hep-ph↗

arXiv2014

Comparing and Counting Logs in Direct and Effective Methods of Resummation

We compare methods to resum logarithms in event shape distributions as they have been used in perturbative QCD directly and in effective field theory. We demonstrate that they are equivalent. In showing this equivalence, we are able to put standard soft-collinear effective theory (SCET) formulae for cross sections in momentum space into a novel form more directly comparable with standard QCD formulae, and endow the QCD formulae with dependence on separated hard, jet, and soft scales, providing potential ways to improve estimates of theoretical uncertainty. We show how to compute cross sections in momentum space to keep them as accurate as the corresponding expressions in Laplace space. In particular, we point out that that care is required in truncating differential distributions at N$^k$LL accuracy to ensure they match the accuracy of the corresponding cumulant or Laplace transform. We explain how to avoid such mismatches at N$^k$LL accuracy, and observe why they can also be avoided by working to N$^k$LL$'$ accuracy.

hep-ph↗

arXiv2014

Study of the 125 GeV Standard Model Higgs Boson Partial Widths and Branching Fractions

The discovery of the Higgs boson, with mass known to better than the percent level, enables for the first time precision Higgs boson analyses. Toward this goal, we define an expansion formalism of the Higgs boson partial widths and branching fractions that facilitates such studies. This expansion yields the observables as a perturbative expansion around reference values of Standard Model input observables (quark masses, QCD coupling constant, etc.). We compute the coefficients of the expansion using state-of-the-art results. We also study the various sources of uncertainties in computing the partial widths and branching fractions more precisely. We discuss the impact of these results with efforts to discern new physics through precision Higgs boson studies.

hep-ph↗

arXiv2012

Does $H \to γγ$ Taste like vanilla New Physics?

We analyse the interplay between the Higgs to diphoton rate and electroweak precision measurements contraints in extensions of the Standard Model with new uncolored charged fermions that do not mix with the ordinary ones. We also compute the pair production cross sections for the lightest fermion and compare them with current bounds.

hep-ph↗

arXiv2012

Three-particle templates for boosted Higgs

We explore the ability of three-particle templates to distinguish color neutral objects from QCD background. This method is particularly useful to identify the standard model Higgs, as well as other massive neutral particles. Simple cut-based analysis in the overlap distributions of the signal and background is shown to provide a significant rejection power. By combining with other discriminating variables, such as planar flow, and several variables that depend on the partonic template, three-particle templates are used to characterize the influence of gluon emission and color flow in collider events. The performance of the method is discussed for the case of a highly boosted Higgs in association with a leptonically-decaying W boson.

hep-ph↗

arXiv2011

Structure of Fat Jets at the Tevatron and Beyond

Boosted resonances is a highly probable and enthusiastic scenario in any process probing the electroweak scale. Such objects when decaying into jets can easily blend with the cornucopia of jets from hard relative light QCD states. We review jet observables and algorithms that can contribute to the identification of highly boosted heavy jets and the possible searches that can make use of such substructure information. We also review previous studies by CDF on boosted jets and its measurements on specific jet shapes.

hep-ph↗

arXiv2011

Two-loop matching factors for light quark masses and three-loop mass anomalous dimensions in the RI/SMOM schemes

Light quark masses can be determined through lattice simulations in regularization invariant momentum-subtraction(RI/MOM) schemes. Subsequently, matching factors, computed in continuum perturbation theory, are used in order to convert these quark masses from a RI/MOM scheme to the MS-bar scheme. We calculate the two-loop corrections in quantum chromodynamics(QCD) to these matching factors as well as the three-loop mass anomalous dimensions for the RI/SMOM and RI/SMOM_gamma_mu schemes. These two schemes are characterized by a symmetric subtraction point. Providing the conversion factors in the two different schemes allows for a better understanding of the systematic uncertainties. The two-loop expansion coefficients of the matching factors for both schemes turn out to be small compared to the traditional RI/MOM schemes. For nf=3 quark flavors they are about 0.6-0.7% and 2%, respectively, of the leading order result at scales of about 2 GeV. Therefore, they will allow for a significant reduction of the systematic uncertainty of light quark mass determinations obtained through this approach. The determination of these matching factors requires the computation of amputated Green's functions with the insertions of quark bilinear operators. As a by-product of our calculation we also provide the corresponding results for the tensor operator.

hep-ph↗

arXiv2010

Template Overlap Method for Massive Jets

We introduce a new class of infrared safe jet observables, which we refer to as template overlaps, designed to filter targeted highly boosted particle decays from QCD jets and other background. Template overlaps are functional measures that quantify how well the energy flow of a physical jet matches the flow of a boosted partonic decay. Any region of the partonic phase space for the boosted decays defines a template. We will refer to the maximum functional overlap found this way as the template overlap. To illustrate the method, we test lowest-order templates designed to distinguish highly-boosted top and Higgs decays from backgrounds produced by event generators. For the functional overlap, we find good results with a simple construction based on a Gaussian in energy differences within angular regions surrounding the template partons. Although different event generators give different averages for our template overlaps, we find in each case excellent rejection power, especially when combined with cuts based on jet shapes. The template overlaps are capable of systematic improvement by including higher order corrections in the template phase space.

hep-ph↗

arXiv2009

Threshold Resummation for Di-hadron Production in Hadronic Collisions

We study the resummation of large logarithmic perturbative corrections to the partonic cross sections relevant for di-hadron production in hadronic collisions, H1 H2 -> h1 h2 X, at high invariant mass of the produced hadron pair. These corrections arise near the threshold for the partonic reaction and are associated with soft-gluon emission. We perform the resummation to next-to-leading logarithmic accuracy, and show how to incorporate consistently cuts in rapidity and transverse momentum of the observed particles. We present numerical results for fixed-target and ISR regimes and find enhancements over the next-to-leading order cross section, which significantly improve the agreement between theoretical predictions and data.

hep-ph↗

arXiv2008

Top Jets at the LHC

We study the reconstruction of high p_T hadronically-decaying top quarks at the LHC. The main challenge in identifying energetic top quarks is that the decay products become increasingly collimated. This reduces the efficacy of conventional methods that exploit the topology of the top decay chain. We focus on the cases where the decay products of the top quark are reconstructed as a single jet, a "top-jet". The most basic "top-tag" method based on jet mass measurement is considered in detail. To analyze the top-tagging method, theoretical and experimental aspects of the QCD jet background are examined. Based on QCD factorization, we derive a simple analytic approximation for the shape of the QCD jet mass spectrum. We observe a good agreement with the Monte Carlo simulation. We consider high-p_T t\bar{t} production in the Standard Model as an example, and show that our theoretical QCD jet mass distributions can efficiently characterize the background via sideband analyses. We show that with 25 fb^{-1} of data, our approach allows us to resolve top-jets with p_T $\ge$ 1 TeV from the QCD background, and about 1.5 TeV top-jets with 100 fb^{-1}, without relying on b-tagging. To further improve the significance we consider jet shapes (analyzed in 0807.0234 [hep-ph]), which resolve the substructure of energy flow inside cone jets. A method of measuring top quark polarization by using the b-quark p_T is also presented. The main advantages of our approach are: (i) the mass distributions are driven by first principle calculations, instead of relying solely on Monte Carlo simulation; (ii) for high p_T jets (p_T $\ge$1 TeV), IR-safe jet shape variables are robust against detector resolution effects. Our analysis is applicable to other boosted massive particles such as the electroweak gauge bosons and the Higgs.

hep-ph↗

arXiv2008

Threshold Resummation for the Top Quark Charge Asymmetry

We study the QCD charge asymmetry in t\bar{t} production at the Tevatron. We investigate the role of higher orders in perturbation theory by considering the resummation of potentially large logarithmic corrections that arise near partonic threshold. This requires us to employ the rapidity-dependent anomalous dimension matrices that describe color mixing due to soft gluon emission in both quark- and gluon-initiated processes. The charge asymmetry appears directly in the resummed cross section at next-to-leading logarithm (NLL), and we find that the first-order expansion of the NLL resummed charge asymmetry reproduces the known fixed-order result for the asymmetry well. Beyond its lowest order, the asymmetric component of the cross section is enhanced by the same leading-logarithmic threshold corrections as the total cross section. As a result, the charge asymmetry is robust with respect to the higher-order perturbative corrections generated by threshold resummation. We observe that the asymmetry increases with pair mass and with scattering angle.

hep-ph↗

arXiv2008

Substructure of high-p_T Jets at the LHC

We study high-pt jets from QCD and from highly-boosted massive particles such as tops, W, Z and Higgs, and argue that infrared-safe observables can help reduce QCD backgrounds. Jets from QCD are characterized by different patterns of energy flow compared to the products of highly-boosted heavy particle decays, and we employ a variety of jet shapes, observables restricted to energy flow within a jet, to explore this difference. Results from Monte Carlo generators and arguments based on perturbation theory support the discriminating power of the shapes we refer to as planar flow and angularities. We emphasize that for massive jets, these and other observables can be analyzed perturbatively.

hep-ph↗