10th International Workshop on Boosted Object Phenomenology, Reconstruction and Searches in HEP (BOOST 2018)

16 Jul 2018, 08:30 → 20 Jul 2018, 18:00 Europe/Paris

Charpak Amphitheater (Paris)

Gregory Soyez (IPhT, CEA Saclay), Matteo Cacciari (LPTHE Paris), Reina Coromoto Camacho Toro (University of Chicago (US))

BOOST 2018 is the tenth of a series of successful joint theory/experiment workshops that bring together the world's leading experts from theory and LHC experiments to discuss the latest progress and develop new approaches on the reconstruction of and use of boosted decay topologies in order to search for new physics.

This year, the conference is hosted in Paris, on the Jussieu campus of Sorbonne Université, and it will mark the tenth anniversary of the BDRS paper that bootstrapped the boosted jets field. The talks will take place in the Charpak amphitheater located on the campus.

Previous editions took place in SLAC (2009), Oxford (2010), Princeton (2011), Valencia (2012), Flagstaff (2013), London (2014), Chicago (2015), Zurich (2016) and Buffalo (2017).

The conference will cover the following topics:

• Searches using jet substructure
• New jet substructure algorithms
• Measurements and modelling
• First-principles calculations
• Machine learning
• Pileup mitigation
• Heavy-ion collisions
• Future colliders

poster.pdf

Monday, 16 July

08:30 → 10:30 Boost Camp

08:30 Boost Camp

Boost Camp is an introduction to the conference made primarily for new people to the field.

Speakers: Laís Sarem Schunk (IPhT, CEA - Saclay), Steven Schramm (Universite de Geneve (CH))

boost2018.pdf

BOOST Camp 2018.pdf

10:30 → 11:00 Coffee break

11:00 → 12:20 Introduction

11:00 Advances in jet substructure for tagging

Speaker: Roman Kogler (Hamburg University (DE))

BOOST18_jetsubstructure_for_tagging.pdf

11:40 Advances in jet substructure for precision physics and QCD

Speaker: Ben Nachman (University of California Berkeley (US))

Nachman_BOOST2018.pdf

12:20 → 14:00 Free lunch

14:00 → 16:05 Performance

14:00 Jet and Missing Momentum Reconstruction and Calibration in ATLAS

Jets form one of the most important quantities to reconstruct at a hadron collider and many searches and measurements are dominated by the uncertainties on the jet energy scale. The procedure of calibrating ATLAS Anti-kt R=0.4 jets using in situ techniques and extrapolation to other radii is presented. Missing transverse momentum (MET) is used to infer the presence of high transverse momentum neutrinos or other weakly interacting neutral particles. The reconstruction and performance in data and simulation of the MET obtained with different class of jets including Particle Flow jets and different pile-up suppression schemes are presented. The performance of a new definition of the MET based on the resolutions of the individual objects to form an estimator for the MET resolution is also shown.

Speaker: Christopher Young (CERN)

Presentation.pdf

14:25 Jet+MET performance in CMS

The talk will report on the most recent studies concerning the performance of jet and missing transverse energy reconstruction at CMS

Speaker: Viola Sordini (Centre National de la Recherche Scientifique (FR))

JetMET_CMS_Boost18_VSordini.pdf

14:50 Reconstruction and Calibration of Large-R jets and In-situ techniques for their Calibration

Searches for new heavy resonances, supersymmetric particles, heavy vector quarks, and measurements of highly boosted Standard Model processes increasingly rely on large radius jets to reconstruct hadronically decaying boosted objects. The reconstruction of these along with their calibration of both mass and transverse momentum will be discussed. In situ techniques are then used to derive corrections and uncertainties, on both the energy scale and the jet properties.

Speaker: Ayana Tamu Arce (Duke University (US))

largeR_reco_insitu_arce.pdf

15:15 Triggering boosted objects at 25 ns at CMS

With the High-Luminosity (HL) LHC ahead, bringing the discrimination power of jet-substructure-based jet tagging algorithms to Level 1 trigger selection will be a major asset to guarantee an optimal usage of the limited trigger resources at downstream steps. To do so, the CMS experiment is intensively investigating the possibility of running the particle-flow (PF) reconstruction algorithm on low-latency FPGA cards, part of the L1 hardware electronics, taking advantage of the advent of track reconstruction at L1. The availability of particle candidates at L1 opens the possibility to compute jet substructure observables and develop new boosted jet tagging algorithms to improve trigger and data acquisition performance, with great advantage for HLT and offline selections downstream, and eventually for physics analysis. This technological revolution will make it possible to use advanced pileup mitigations techniques, such as pileup per particle identification (PUPPI), already in the L1 reconstruction, reconceptualising jet algorithms for L1 trigger at hadron colliders. We present proof-of-principle studies on both physics and hardware performance of prototype jet algorithms foreseen by CMS for the HL-LHC.

Speaker: Jennifer Ngadiuba (CERN)

PFatL1Trigger_BOOST2018.pdf

15:40 Inputs, grooming and track assisted techniques for jet substructure

In order to mitigate the effect of pile-up and to improve the resolution of substructure variables for highly boosted hadronically decaying particles various grooming techniques and jet inputs are evaluated. Track assisted techniques are also introduced to improve over the calorimeter limited granularity in the highest boosted regime. Boosted Higgs bosons decaying into two collimated b-jets are reconstructed in a single large radius jet with variable radius subjets. The performance of these techniques is presented in ATLAS.

Speaker: Sanmay Ganguly (Weizmann Institute of Science (IL))

Boost_Talk_Ganguly.pdf

16:05 → 16:35 Coffee break

16:35 → 18:15 Calculations

16:35 Improving the Simulation of Quark and Gluon Jets with Herwig 7

The properties of quark and gluon jets, and the differences between them, are increasingly important at the LHC. However, Monte Carlo event generators are normally tuned to data from e+e− collisions which are primarily sensitive to quark-initiated jets. In order to improve the description of gluon jets we make improvements to the perturbative and the non-perturbative modelling of gluon jets and include data with gluon-initiated jets in the tuning for the first time. The resultant tunes significantly improve the description of gluon jets and are now the default in Herwig 7.1. The talk will be base on our recent publication Eur.Phys.J. C77 no.12, 876.

Speaker: Andrzej Konrad Siodmok (Polish Academy of Sciences (PL))

qgBoostSiodmok.pdf

17:00 Quark jet rates and quark gluon discrimination in multi-jet final states

We calculate quark and gluon jet fraction in multi-jet final states at the LHC, which is based on perturbative QCD at next-to-double logarithmic accuracy. We find a measurable scaling pattern of the fraction. This is related to a performance of new physics searches using quark-gluon jet discrimination in multi-jet final states, and would be useful for more understanding of QCD and tuning of Monte-Carlo generators. We also introduce a variable related to jet flavors in multi-jet final states, and propose a data-driven method using the variable to improve the performance of BSM searches and to reduce systematic uncertainties of the analysis. We show how the background rejection increase for signals which produce many quark jets.

Speaker: Dr Yasuhito Sakaki (KAIST)

BOOST2018_SAKAKI.pdf

17:25 How the parton shower cutoff affects the heavy quark mass parameter

We show that using an infrared cutoff in the parton shower evolution for massive quarks implies that one employs a short-distance mass scheme, i.e. a mass scheme that does not have the O(Lambda_QCD) renormalon contained in the pole mass. Our analysis considers an angular ordered shower based on coherent branching and is based on jet masses. In this context we determine explicitly the relation of this short-distance mass to the pole mass. The basis of our analysis is (a) an analytic solution of the parton shower evolution and (b) that the infrared cut of the angular ordered shower can be implemented into analytic calculations in the framework of SCET. Numerical comparison to the Herwig event generator confirms our analytic results. The outcome of our analysis proves that the top quark mass parameter contained in multi-purpose event generators is in general not the pole mass.

Speaker: Daniel Samitz

BOOST_Samitz.pdf

17:50 Infrared Safety of a Neural-Net (Machine Learning) Top Tagging Algorithm

Neural network-based algorithms provide a promising approach to jet classification problems, such as boosted top jet tagging. In this talk, I will discuss that the jet observable defined by the convolutional neural network obeys the canonical definition of infrared safety: it is unaffected by the presence of the extra gluon, as long as it is soft or collinear with one of the quarks, which indicates that the convolutional neural network tagger is robust with respect to possible mis-modeling of soft and collinear final-state radiation by Monte Carlo generators.

Speaker: Seung Joon Lee (Korea University)

BOOST2018_SLEE.pdf

18:30 → 21:30 Cocktail

Tuesday, 17 July

09:00 → 10:40 Performance

09:00 Pileup mitigation for jets with Constituent Subtraction

Pileup has a significant impact on jet reconstruction and on the ability to identify hadronically decaying boosted objects. The Constituent Subtraction is a pileup mitigation method which operates at the level of jet or event constituents and which is successfully used by experiments at the LHC. This talk explores its application on the event constituents before jet clustering. Several improvements are presented, along with performance studies for expected pileup conditions during the LHC Run 2 and Run 3. An overview of the use of the Constituent Subtraction in experiments at the LHC is also provided.

Speaker: Peter Berta (Johannes Gutenberg Universitaet Mainz (DE))

PeterBerta_CS_17.7.2018.pdf

09:25 Pileup mitigation in CMS

The talk will report on the studies and techniques developed at CMS for the pileup mitigation.

Speaker: Anna Benecke (Hamburg University (DE))

boost2018_benecke.pdf

09:50 Tagging of Hadronically Decaying Particles and the Determination of boosted tagger efficiencies in ATLAS

Tagging algorithms that discriminate between massive large-R QCD jets and large-R jets that capture the decays of W/Z boson or top quarks are developed. The methods of tagging these objects are described. In order to improve the boosted taggers performance and the precision on the tagging efficiency and rejection new in-situ techniques are applied. The precision obtained by applying the in-situ techniques is presented. Performance of machine learning taggers are also presented.

Speaker: Andreas Sogaard (University of Edinburgh (GB))

BOOST2018_AndreasSogaard.pdf

10:15 Tagger performance in data

The talk will report on most recent results on tagger performance in data by the CMS collaboration

Speaker: Cristina Ana Mantilla Suarez (Johns Hopkins University (US))

CMS_TaggerPerf_BOOST2018.pdf

10:40 → 11:10 Coffee break

11:10 → 12:25 Measurements

11:10 Jet substructure measurements at ALICE

Jet substructure measurements at ALICE
Leticia Cunqueiro, for the ALICE Collaboration

The yields and fragmentation pattern of jets are modified by the dense medium created in heavy ion collisions relative to proton proton collisions. The main scope of the jet program in heavy ions is to address those modifications, relate them to specific aspects of the dynamics of the jet-medium interactions and to extract a microscopic description of the medium.
Techniques like grooming, iterative de-clustering and mapping the jet tree branches onto Lund maps can help to individuate regions of phase space dominated by the different components of the experimental background and expected signal, such as combinatorial background, medium response, vacuum radiation, multiple soft or semi-hard medium-induced radiation. We will discuss recent jet substructure measurements in Pb-Pb collisions at ALICE.

Speaker: Leticia Cunqueiro Mendez (Oak Ridge National Laboratory - (US))

Boost2018Cunqueiro.pdf

11:35 Modifications to Jet Spectra and Substructure in PbPb Collisions with CMS

Study of inclusive jet spectra and substructure are powerful tools to study modifications of the parton shower as it propagates through the hot and dense medium produced in PbPb collisions. The observed suppression of jet production and the modification to jet substructure results in strong constraints on jet quenching mechanisms in heavy ion collisions. Combining substructure measurements with a boson tag can provide an estimate of the initial parton energy before modification in the medium. In this talk, a radius scan of jet nuclear modification factor at high pt out to R=1.0 will be reported, in addition to the latest results inclusive and boson-tagged jet substructure. These results map the loss of jet energy outside of the cone in a theoretically controlled way.

Speaker: Christopher Mc Ginn (Massachusetts Inst. of Technology (US))

boost2018_5TeVCMSJet_CFM_20180710.pdf

12:00 Comparison of jet mass in $pp$ and Pb+Pb collisions using ATLAS at $\sqrt{s_{NN}}$=5.02 TeV

The mass of inclusive jets in lead-lead and proton-proton collisions at $\sqrt{s_{\mathrm{NN}}}=5.02$ TeV is reported using the ATLAS Run 2 data at the LHC. Jet substructure observables are of interest to access the internal structure of jets produced in proton-proton collisions. These observables are sensitive to the angular and momentum correlations of the jet fragments, and are thus useful in characterizing the modification of jets in heavy ion collisions providing complementary information to single particle fragmentation functions. In this analysis, the mass of anti-k_{t} R = 0.4 jets is measured as a function of centrality and jet transverse momentum in Pb+Pb collisions and compared to measurements at the same collision energy in proton-proton collisions. The current status of this measurement is presented and discussed in the context of other jet measurements in Pb+Pb collisions.

Speaker: Anne Marie Sickles (Univ. Illinois at Urbana Champaign (US))

sickles_071718.pdf

12:25 → 13:45 Organised Lunch

13:45 → 14:30 Poster session

Sung_Hak_Lim BOOST_Poster.pdf BOOST_Supplementary.pdf

14:30 → 16:10 Machine Learning

14:30 Novel Jet Observables from Machine Learning

Ref [1] presented a method to construct novel observables exclusively from information identified by the machine, by using a parametrization of the N-body phase space coordinates at the point of saturation of discrimination power. We have now studied how to extend this approach in an automated way to higher N-body phase space. We use boosted Z' vs QCD discrimination as a testing ground to present a machine learning framework that adversarially carries out the task of developing the optimal discriminant for the problem. Following on from a long string of results that have improved upon previous N-prong taggers, this technique has important implications for future development of optimal discrimination observables.

[1] K. Datta and A. Larkoski, Novel Jet Observables from Machine Learning, JHEP 03, 086 (2018)

Speaker: Kaustuv Datta (ETHZ - ETH Zurich)

BOOST_2018_Datta.pdf

BOOST_2018_Datta.pptx

14:55 How to train taggers on data

In the current era of high energy experiments, we are faced with an overwhelming amount of data and the limiting uncertainty in new physics searches can often come from theory and not experiment. In our efforts to develop new approaches to extract complex signals from large backgrounds, BDTs, neural networks and other machine learning techniques are becoming increasingly significant. These tools allow us to find patterns in data that would be impossible with a simple cut-and-count approach. In this work we use deep learning tools to develop and validate a new physics tagger that can be trained on data and does not rely on specific models of new physics.

Speaker: Jennifer Thompson (ITP Heidelberg)

tagging_from_data.pdf

15:20 Learning to Classify from Impure Samples with High-Dimensional Data

Machine learning in high energy physics relies heavily on simulation for fully supervised training. This often results in sub-optimal classification when ultimately applied to (unlabeled) data. In addition to describing a new method for weak supervision (learning directly from data) called Classification Without Labels (CWoLa), we show for the first time how to apply these techniques to high-dimensional data, where significant architectural changes are required. This is critically important for learning from and about the full radiation pattern inside jets.

Speakers: Ben Nachman (University of California Berkeley (US)), Eric Metodiev (Massachusetts Institute of Technology), Patrick Komiske (Massachusetts Institute of Technology)

WeakSupervision_BOOST2018.pdf

15:45 CWoLa Hunting

Classification Without Labels (CWoLa) is a Machine Learning strategy which can be used to classify event categories (e.g. quark jet vs gluon jet, or BSM signal vs SM background) starting from mixed event samples, which are inevitable at particle collider experiments. I will illustrate how this strategy can be used to uncover BSM resonance signals which would otherwise be completely buried under large SM backgrounds by taking advantage of as much information as possible in the events, focusing on a toy example where a di-fat jet resonance with unusual jet substructure is hidden amongst standard model dijets. This strategy can be applied directly to data without a requirement for specific signal models, simulated signal or background Monte Carlo events, or carefully chosen cuts.

Speaker: Jack Collins (University of Maryland and Johns Hopkins University)

BOOST_CWoLa_Hunting.pdf

16:10 → 16:40 Coffee break

16:40 → 17:55 Measurements

16:40 Measuring jet substructure observables at the ATLAS Experiment

Jet substructure observables have significantly extended the search program for physics beyond the Standard Model at the Large Hadron Collider. The state-of-the-art tools have been motivated by theoretical calculations, but there has never been a direct comparison between data and calculations of jet substructure observables that are accurate beyond leading-logarithm approximation. Such observables are significant not only for probing the collinear regime of QCD that is largely unexplored at a hadron collider, but also for improving the understanding of jet substructure properties that are used in many studies at the Large Hadron Collider. The ATLAS collaboration has recently performed several measurements of precision jet substructure at 13 TeV that will significantly extend our understanding of both the perturbative and non-perturbative aspects of jet formation. These measurements of jet mass in various topologies as well as other properties of jet fragmentation such as charged-particle multiplicity and the properties of gluon splitting to bottom quarks are unfolded to correct for detector effects and compared with a variety of predictions.

Speaker: Jennifer Kathryn Roloff (Harvard University (US))

jssMeasurementsAtlas.pdf

17:05 Measurements using jet substructures for SM physics at CMS

The talk will report on measurements using jet substructures for top-quark physics and other standard-model channels by the CMS collaboration

Speaker: Matthias Ulrich Mozer (KIT - Karlsruhe Institute of Technology (DE))

mbi_mmozer.pdf

17:30 Measurements with highly boosted top quarks using the ATLAS detector

The high energy of the LHC allows access to large numbers of high transverse momentum top quarks. Measurements of differential cross-sections in top quark pair production at 13 TeV with the ATLAS detector are presented. They are performed using the lepton+jets and all-hadronic final states. Jet substructure techniques are used to identify hadronically decaying top quarks. The measurements are corrected for detector effects to obtain differential cross-sections at particle-level in a fiducial region close to the event selection. These measurements probe our understanding of top quark pair production in the TeV regime. The results, unfolded to particle and parton level, are compared to predictions of Monte Carlo generators implementing NLO matrix elements matched with parton showers and NNLO QCD theory calculations.

Speaker: Walter Hopkins (University of Oregon (US))

diffXSecBoost2018_Hopkins.pdf

Wednesday, 18 July

09:00 → 10:40 Calculations

09:00 Nonperturbative Effects in Groomed Event Shapes

Jet grooming has an important significance in the study of QCD event shapes. Soft drop grooming is a great tool for reducing the soft contamination to the jet, while at the same time allowing one to calculate groomed event shape cross sections in perturbation theory. It also results in reduced corrections to the perturbative cross section from hadronization and underlying event. However, despite these advantages, due to complicated nature of the algorithm, it is a non-trivial problem to include consistently a description of non-perturbative effects. This is absolutely essential if one aims to use groomed observables to carry out measurements, such as $\alpha_s$, where an appropriate parameterization of the non-perturbative effects is needed. In this talk I address these issues, deriving results solely from field theory wherever possible.

Speaker: Dr Aditya Pathak (University of Vienna)

Adi_Boost18.pdf

09:25 Calculating the Pull Angle

The pull angle was introduced as a sensitive probe of the color flow between jets and has been measured in W boson decays by ATLAS. Despite being infrared and collinear unsafe, it can be calculated in resummed perturbation theory. In this talk, I will review the pull angle, introduce our calculation, and compare it to ATLAS data.

Speaker: Andrew Larkoski (Reed Collge)

BOOST_2018_Larkoski.pdf

09:50 Top taggers : an analytical perspective

We carry out first analytical calculations for top taggers along similar lines to those already performed for the most common W/Z/H taggers. We discuss both the CMS top tagger as well as the Y-splitter method adapted for top tagging. A novel feature of our calculations is the use of triple-collinear splitting functions to describe 1 \to 3 decays. We find that the default CMS tagger is infrared unsafe and propose new variants of the tagger that are more robust theoretically, while giving a near identical performance to the default tagger. We also show that the Y-splitter based methods give rise to a much stronger suppression of the QCD background but that eventual tagger performance also depends crucially on details of the Sudakov form factor for the coloured signal jets. Lastly we compare analytics to Monte Carlo and draw some conclusions about factors influencing tagger performance and stability.

Speaker: Marco Guzzi (Southern Methodist University)

mguzzi-boost2018.pdf

10:15 Soft Gluons and Non-global Observables

We present an evolution algorithm for soft gluon exchanges and the resummation of non-global logarithms. Our approach applies to generic hard-scattering processes involving any number of coloured partons and we present a reformulation of the algorithm in such a way as to make the cancellation of infrared divergences explicit. Handling large colour matrices presents the most significant challenge to numerical implementations and we present a means to expand systematically about the leading colour approximation. We present first numerical results obtained with a framework which can cover a large class of such evolution algorithms.

Speaker: Prof. Mike Seymour (University of Manchester)

Boost2018Seymour.pdf

Boost2018Seymour.pptx

10:40 → 11:10 Coffee break

11:10 → 12:25 Searches

11:10 ATLAS Searches using Jet Substructure Techniques

Jet Substructure (JSS) is a critical tool in the search for new physics at the energy frontier. The large center-of-mass energy of the LHC can produce heavy new particles which subsequently would decay into very energetic SM particles. Without JSS techniques, our ability to identify decays to Vector Bosons, Higgs Bosons, or top-quarks would decrease dramatically. Also, to evade the restrictions of the online trigger systems, JSS techniques are also employed to search for light new particles produced with large transverse momentum. This talk will give an overview of the ATLAS Run 2 search program where JSS is of the greatest importance.

Speaker: Francesco Guescini (TRIUMF (CA))

20180718.v10.pdf

11:35 Highlights on searches for new physics with vector bosons and higgs bosons in boosted topologies

The talk will report on highlights on searches for new physics with vector bosons and higgs bosons in boosted topologies at the CMS experiment

Speaker: Thea Aarrestad (Universitaet Zuerich (CH))

BoostedSearches_BOOST18.pdf

12:00 Stealth bosons and where to find them

I want to discuss jets resulting from the cascade decay of a new particle into four quarks, focusing on: (i) their substructure, which looks like QCD for standard taggers (ii) their groomed mass, which often differs considerably from the true mass of the originating particle. I also want to discuss how to distinguish such jets from the QCD background.

Speaker: Juan Antonio Aguilar-Saavedra (University of Granada)

SBAWTFD.pdf

12:25 → 14:00 Organised lunch (lunchbox)

14:00 → 14:50 Searches

14:00 Search for supersymmetry using dedicated jet techniques with the ATLAS detector

The hunt for supersymmetry at the LHC continues to probe increasingly challenging experimental final states. This talk presents recent ATLAS results that utilise boosted jet techniques, explore all-hadronic final states while searching for signs of small supersymmetric signals, or map out the landscape for non-prompt decays of heavy supersymmetric particles.

Speaker: David Miller (University of Chicago (US))

Miller_BOOST2018_BoostedSUSY_18Jul2018.pdf

14:25 Highlights on searches for new physics with top quarks in boosted topologies

The extension of sensitivity for strongly-produced supersymmetry (SUSY) to increasingly higher masses makes the reconstruction and identification of boosted objects to an essential tool for current and future SUSY searches. These objects can originate from the final stage of a short decay chain, or as heavy bosons produced in a decay chain. The talk summaries the use of large-radius jets and substructure techniques in searches such as the ones for the pair production of gluinos or third generation squarks in proton-proton collisions at 13 TeV.

Speaker: Christine Angela Mc Lean (SUNY Buffalo)

BoostedTopSearches_McLean.pdf

14:50 → 15:40 Future Colliders

14:50 Boosted top quarks at multi-TeV CLIC

The Compact Linear Collider (CLIC) is a proposed high-luminosity linear electron-positron collider at the energy frontier. For optimal physics potential, CLIC will be operated in a staged approach, with three energy stages ranging from a few hundred GeV up to an ultimate centre-of-mass energy of 3 TeV. At the higher energy stages, a large fraction of the top quarks are produced with significant boosts resulting in a highly collimated jet environment. In this talk, we present recent results on the capability of the CLIC detector to identify highly boosted top quarks by exploiting the internal sub-structure of the final state jets. These methods play an important role in extending the top quark physics program to higher energies, which is demonstrated in a study of semi-leptonic top-quark pair production. The results presented are based on detailed Monte Carlo simulation studies including full detector simulation and relevant background processes.

Speaker: Lars Rickard Strom (CERN)

boost2018_stroem.pdf

15:15 Jet observables and stops at 100 TeV collider

A future proton-proton collider with center of mass energy around 100 TeV will have a remarkable capacity to discover massive new particles and continue exploring weak scale naturalness. In this work we will study its sensitivity to two stop simplified models as further examples of its potential power: pair production of stops that decay to tops or bottoms and higgsinos; and stops that are either pair produced or produced together with a gluino and then cascade down through gluinos to the lightest superpartner (LSP). In both simplified models, super-boosted tops or bottoms with transverse momentum of order TeV will be produced abundantly and call for new strategies to identify them. We will apply a set of simple jet observables, including track-based jet mass, N-subjettiness and mass drop, to tag the boosted hadronic or leptonic decaying objects and suppress the Standard Model as well as possible SUSY backgrounds. Assuming 10% systematic uncertainties, the future 100 TeV collider can discover (exclude) stops with masses up to 6 (7) TeV with 3 inverse attobarns of integrated luminosity if the stops decay to higgsinos. If the stops decay through gluinos to LSPs, due to additional SUSY backgrounds from gluino pair production, a higher luminosity of about 30 inverse attobarns is needed to discover stops up to 6 TeV. We will also discuss how to use jet observables to distinguish simplified models with different types of LSPs. The boosted top or bottom tagging strategies developed in this paper could also be used in other searches at a 100 TeV collider. For example, the strategy could help discover gluino pair production with gluino mass close to 11 TeV with 3 inverse attobarns of integrated luminosity.

Speaker: Shing Chau Leung (Brown University)

100TeVBoost18.pdf

15:40 → 16:10 Coffee break

16:10 → 17:25 New Tools

16:10 Energy Flow and Jet Substructure

I will discuss a set of observables, the Energy Flow Polynomials (EFPs), which form a complete, linear basis for IRC-safe observables. I will demonstrate that, on the problems of quark/gluon discrimination, boosted W tagging, and boosted top tagging, the performance of EFPs with linear classification is comparable to that of complex, modern machine learning techniques. Efficient computation of the EFPs is vital to their practical use, and I will develop novel algorithms that make use of the graph-theoretic interpretation of EFPs to improve their computational complexity over that of an arbitrary N-particle energy correlator. Finally, I will introduce the Energy Flow Moments (EFMs), tensors that allow for all EFPs to be computed in time linear in the multiplicity, and discuss how EFMs can be used to understand redundancies in the energy flow basis.

Speaker: Patrick Komiske (Massachusetts Institute of Technology)

PTK_BOOST2018_compressed.pdf

16:35 The Lund Jet Plane

Lund diagrams, a representation of the phase space within jets,
have long been used in discussing parton showers and
resummations. We point out here that they can also serve as a
powerful tool for experimentally characterising the radiation
pattern within jets. We briefly comment on some of their
analytical properties and highlight their scope for constraining
Monte Carlo simulations. We then examine the use of the Lund
plane for boosted electroweak boson tagging. When used as an
input to deep-learning methods it yields high performance.
Furthermore, much of that performance can be reproduced by using
the Lund plane as an input to simpler log-likelihood type
discriminators. This suggests a potential for unique insight and
experimental validation of the features being used by
machine-learning approaches. In the context of our discussion, we
also highlight the importance of accounting for detector effects
when considering the performance of machine-learning approaches.

Speaker: Frederic Alexandre Dreyer (MIT)

talk_boost18.pdf

17:00 New tagging approaches

The talk will report on new tagging approaches developed by the CMS collaboration

Speaker: Loukas Gouskos (Univ. of California Santa Barbara (US))

boost2018-newtaggingcms-LG.pdf

17:25 → 18:35 Meeting of the international committee

Conveners: Peter Loch (University of Arizona (US)), Salvatore Rappoccio (The State University of New York SUNY (US))

17:25 Boost 2021 - Seoul proposal

2021-seoul.pdf

17:35 Boost 2021 - Berkeley proposal

2020-21-LBNL.pdf

17:45 Boost 2020 - Cracow proposal

2020-cracow.pdf

17:55 Boost 2020 - Hamburg proposal

2020-Hamburg.pdf

18:05 Boost 2020 - Manchester proposal

2020-Manchester.pdf

18:15 Boost 2020 - Brighton proposal

2020-BoostBrighton.pdf

18:25 Booston 2019 - status report

2019-Booston.pdf

Thursday, 19 July

09:00 → 10:40 New Tools

09:00 Prospects for a measurement of the W boson mass in the all-jets final state at hadron colliders

Precise measurements of the mass of the $W$ boson are important to test the overall consistency of the Standard Model of particle physics. The current best measurements of the $W$ boson mass come from single production measurements at hadron colliders in its decay mode to a lepton (electron or muon) and a neutrino and pair production of $W$ bosons at lepton colliders, where also the decay mode of the $W$ boson to a quark anti-quark pair has been considered. In this study, prospects for a measurement of the $W$ boson mass in the all-jets final state at hadron colliders are presented. Compared to the lepton plus neutrino final state, a measurement in the all-jets final state could avoid experimental systematic uncertainties related to the measurement of the missing transverse momentum and the theoretical uncertainties related to the transverse mass. Compared to other methods for measuring the $W$ mass, a measurement in the all-jets final state would be complementary in methodology and have systematic uncertainties orthogonal to previous measurements. We have estimated the main experimental and theoretical uncertainties affecting a measurement in the all-jets final state making use of jet substructure techniques.

Speaker: Andreas Hinzmann (University of Hamburg)

180719 - W mass all-jets.pptx.pdf

Link to arxiv

09:25 Jet SIFT-ing

We introduce a new jet clustering algorithm (SIFT: Scale-Invariant Filter Tree), which does not impose a fixed cone size or associated scale on the event. The proposed construction maintains excellent object discrimination for very collimated partonic systems, while asymptotically recovering favorable behaviors of the standard anti-KT algorithm. It is intrinsically suitable (without secondary declustering) for the tagging of highly boosted objects, and applicable to the study of jet substructure. Additionally, it is resilient to pileup, via a concurrent filter on soft wide-angle radiation applied within the primary clustering phase.

Speaker: Prof. Joel Walker (Sam Houston State University)

Jet-SIFTing_Boost.pdf

Movies

• 1_PP_TTB_P0_ORIG.mov
• 2_LEP_TTB_2500_AK5_GST.mov
• 3_LEP_TTB_2500_GST.mov
• 4_PP_TTB_P0_SIFT.mov
• 5_LEP_TTB_2500_GST_SIFT.mov
• clustering_13TeV_pup0.mov
• clustering_13TeV_pup10.mov
• clustering_13TeV_pup20.mov

09:50 On the Topic of Jets

I discuss a new framework to identify underlying classes of jets directly from data without input from simulation or theory. Due to a mathematical connection between mixed samples of jets and emergent themes in documents, statistical methods from topic modeling can be used to extract "jet topics" from data. I apply the jet topics method to extract quark and gluon multiplicity distributions from simulated Z+jet and dijet samples. I compute the jet topics to leading logarithmic accuracy for jet mass and perturbative multiplicity and discuss the relevance for defining quarks and gluons. I also discuss the potential to directly adapt any machine-learned jet tagger as a jet fraction and topic extractor.

Speaker: Eric Metodiev (Massachusetts Institute of Technology)

BOOST2018_Topics_EMM.pdf

10:15 Telescoping deconstruction

I will discuss the telescoping deconstruction (TD) framework and its applications in many contexts, including boosted particle tagging, jet grooming and heavy ion jet modification studies. TD represents jet substructure information using a fixed-order subjet expansion with multiple angular resolutions. Specifically, using TD I will show the dominant feature of color-singlet jet isolation for boosted electroweak boson jets and its analytic understanding using soft-collinear effective theory. I will discuss a class of collinear-drop observables and their use in non-perturbative and jet modification studies. Also, I will show how QCD splitting function can be probed by TD subjet kinematics, and how TD variables can help reveal characteristic features of jet modifications in heavy ion collisions.

Speaker: Dr Yang-Ting Chien (Massachusetts Institute of Technology)

BOOST2018_Chien.pdf

10:40 → 11:10 Coffee break

11:10 → 12:25 Machine Learning

11:10 Study of Recursive Neural Networks in Jet Tagging

We would like to present a study on the behavior of Recursive Neural Networks (RecNNs) in jet tagging. The RecNNs embed jet clustering history recursively to include all the information within a jet. We examine its behavior in different jet tagging tasks, and analyze the connection between RecNNs and the underlying physics.

Speaker: Dr Taoli Cheng (University of Chinese Academy of Sciences)

CHENG_BOOST2018.pdf

11:35 JUNIPR: a Framework for Unsupervised Machine Learning in Particle Physics

In applications of machine learning to particle physics, there is a persistent tension between interpretability and performance. In this paper, this tension is allayed by introducing a novel framework for unsupervised machine learning in particle physics, in which the neural network architecture is built as a scaffolding around a leading-order description of the physics under study. This approach not only reduces the complexity and increases the efficiency of the machine learning models it inspires, but it can also lead to interpretable models, as opposed to deep black boxes. In particular, we present JUNIPR, a framework for studying "Jets using UNsupervised Interpretable PRobability models". Within this framework, we use a deep neural network to construct a probability model for jet physics, i.e. a function that computes the relative differential cross section of individual jets in a sample. Jets at colliders are defined by sequentially clustering final-state particles together; the resulting tree structure augmenting each jet provides the scaffolding for the deep neural network architecture in the JUNIPR framework, enabling predictions that are easy to visualize and interpret. Although neural network architectures in the JUNIPR framework leverage a user-specified sequential tree structure for jets, training such models is unsupervised and unrestricted: the network could decide that the chosen tree structure has little to do with the training data. To test this, both physically-motivated and unphysical trees are considered. JUNIPR-based probability models are shown to perform powerful discrimination through the statistically optimal likelihood-ratio test, and to permit visualizations of this discrimination power at each branching in a jet's tree. Samples from such probability models can also be drawn, providing a data-driven Monte Carlo generator for computing arbitrary physical observables. It is further demonstrated that JUNIPR-based models can efficiently re-weight jets from one (e.g. simulated) data set to agree with jets from another (e.g. experimental) data set. We elaborate significantly on the inner workings of our approach in an attempt to place a foundation for future work in this novel direction.

Speaker: Christopher Frye (Harvard)

frye-boost-2018.pdf

12:00 Fast inference of deep neural networks in FPGAs for particle physics

Recent results at the Large Hadron Collider (LHC) have pointed to enhanced physics capabilities through the improvement of the real-time event processing techniques. Machine learning methods are ubiquitous and have proven to be very powerful in LHC physics, and particle physics as a whole. However, exploration of the use of such techniques in low-latency, low-power FPGA hardware has only just begun. FPGA-based trigger and data acquisition (DAQ) systems have extremely low, sub-microsecond latency requirements that are unique to particle physics. We present a case study for neural network inference in FPGAs focusing on a classifier for jet substructure which would enable, among many other physics scenarios, searches for new dark sector particles and novel measurements of the Higgs boson. While we focus on a specific example, the lessons are far-reaching. We develop a package based on High-Level Synthesis (HLS) called \hlsfml~to build machine learning models in FPGAs. The use of HLS increases accessibility across a broad user community and allows for a drastic decrease in firmware development time. We map out FPGA resource usage and latency versus neural network hyperparameters to identify the problems in particle physics that would benefit from performing neural network inference with FPGAs. For our example jet substructure model, we fit well within the available resources of modern FPGAs with a latency on the scale of 100~ns.

Speaker: Jennifer Ngadiuba (CERN)

HLS4ML_BOOST2018_Ngadiuba.pdf

12:25 → 14:10 Organised Lunch + Anniversary Cake

14:10 → 15:30 Anniversary

dinner.pdf

14:10 10 years of Boost: experimental perspective

Speaker: Jonathan Butterworth (University College London (UK))

10years-exp.pdf

10years-exp.pptx

14:50 10 years of Boost: theory perspective

Speaker: Gavin Salam (CERN)

Boost18-10years-theory-salam.pdf

18:00 → 22:00 Visit of the Orsay museum and conference dinner

Friday, 20 July

09:30 → 10:00 Coffee bbreak

10:00 → 11:40 Summaries

10:00 Summary: experiments

Speakers: Caterina Vernieri (Fermi National Accelerator Laboratory), Caterina Vernieri (Fermi National Accelerator Lab. (US))

boost-2018-expsummary.pdf

10:50 Summary: theory

Speaker: Jesse Thaler (MIT)

jthaler_Boost2018_Summary.pdf

14:00 → 17:00 Hands-on and tutorials

14:00 FastJet tutorial

Speaker: Salvatore Rappoccio (The State University of New York SUNY (US))

Emacs+Tramp+Docker

Emacs+Tramp+Docker Faster

fastjet-tutorial.pdf

vim_and_docker.txt

14:15 Energy Flow tutorial

Speaker: Patrick Komiske (Massachusetts Institute of Technology)

14:30 Lund Jet Plane tutorial

Speaker: Frederic Alexandre Dreyer (MIT)

tutorial_lund.pdf

14:45 Jet topics tutorial

Speaker: Eric Metodiev (Massachusetts Institute of Technology)

BOOST_EF_demo_instructions.txt

15:00 Hand-on time

16:50 Additional material for self-study

Speaker: Gregor Kasieczka (Hamburg University (DE))

DeepLearningOverview_BOOST18.pdf