Subtracted cumulants: Mitigating large background in jet substructure

Yang-Ting Chien; Daekyoung Kang; Kyle Lee; Yiannis Makris

doi:10.1103/physrevd.100.074030

Zero-loss omega-filtered REM and RHEED on the new Zeiss 912

Proceedings, annual meeting, Electron Microscopy Society of America ◽

10.1017/s0424820100087392 ◽

1991 ◽

Vol 49 ◽

pp. 616-617

Author(s):

J.C.H. Spence ◽

J. Mayer

Keyword(s):

Electron Microscopy ◽

Materials Science ◽

Surface States ◽

Ccd Camera ◽

Dark Field ◽

Ion Pump ◽

Magnetic Imaging ◽

Reflection Electron Microscopy ◽

Diffraction Patterns ◽

Large Background

The Zeiss 912 is a new fully digital, side-entry, 120 Kv TEM/STEM instrument for materials science, fitted with an omega magnetic imaging energy filter. Pumping is by turbopump and ion pump. The magnetic imaging filter allows energy-filtered images or diffraction patterns to be recorded without scanning using efficient parallel (area) detection. The energy loss intensity distribution may also be displayed on the screen, and recorded by scanning it over the PMT supplied. If a CCD camera is fitted and suitable new software developed, “parallel ELS” recording results. For large fields of view, filtered images can be recorded much more efficiently than by Scanning Reflection Electron Microscopy, and the large background of inelastic scattering removed. We have therefore evaluated the 912 for REM and RHEED applications. Causes of streaking and resonance in RHEED patterns are being studied, and a more quantitative analysis of CBRED patterns may be possible. Dark field band-gap REM imaging of surface states may also be possible.

Download Full-text

Explainable AI for ML jet taggers using expert variables and layerwise relevance propagation

Journal of High Energy Physics ◽

10.1007/jhep05(2021)208 ◽

2021 ◽

Vol 2021 (5) ◽

Author(s):

Garvita Agarwal ◽

Lauren Hay ◽

Ia Iashvili ◽

Benjamin Mannix ◽

Christine McLean ◽

...

Keyword(s):

Network Performance ◽

Deep Neural Network ◽

Relevant Information ◽

Jet Substructure ◽

Discrimination Ability ◽

Low Level ◽

Performance Improvements ◽

Intermediate Layers ◽

Explainable Ai ◽

General Method

Abstract A framework is presented to extract and understand decision-making information from a deep neural network (DNN) classifier of jet substructure tagging techniques. The general method studied is to provide expert variables that augment inputs (“eXpert AUGmented” variables, or XAUG variables), then apply layerwise relevance propagation (LRP) to networks both with and without XAUG variables. The XAUG variables are concatenated with the intermediate layers after network-specific operations (such as convolution or recurrence), and used in the final layers of the network. The results of comparing networks with and without the addition of XAUG variables show that XAUG variables can be used to interpret classifier behavior, increase discrimination ability when combined with low-level features, and in some cases capture the behavior of the classifier completely. The LRP technique can be used to find relevant information the network is using, and when combined with the XAUG variables, can be used to rank features, allowing one to find a reduced set of features that capture part of the network performance. In the studies presented, adding XAUG variables to low-level DNNs increased the efficiency of classifiers by as much as 30-40%. In addition to performance improvements, an approach to quantify numerical uncertainties in the training of these DNNs is presented.

Download Full-text

Advances in Jet Substructure at the LHC

10.1007/978-3-030-72858-8 ◽

2021 ◽

Author(s):

Roman Kogler

Keyword(s):

Jet Substructure

Download Full-text

Search for $$ t\overline{t} $$ resonances in fully hadronic final states in pp collisions at $$ \sqrt{s} $$ = 13 TeV with the ATLAS detector

Journal of High Energy Physics ◽

10.1007/jhep10(2020)061 ◽

2020 ◽

Vol 2020 (10) ◽

Author(s):

G. Aad ◽

◽

B. Abbott ◽

D. C. Abbott ◽

A. Abed Abud ◽

...

Keyword(s):

Production Cross ◽

Hadron Collider ◽

Top Quarks ◽

Atlas Detector ◽

Proton Collision ◽

Large Radius ◽

Final States ◽

Jet Substructure ◽

High Transverse Momentum ◽

Hadronic Final States

Abstract This paper presents a search for new heavy particles decaying into a pair of top quarks using 139 fb−1 of proton-proton collision data recorded at a centre-of-mass energy of $$ \sqrt{s} $$ s = 13 TeV with the ATLAS detector at the Large Hadron Collider. The search is performed using events consistent with pair production of high-transverse-momentum top quarks and their subsequent decays into the fully hadronic final states. The analysis is optimized for resonances decaying into a $$ t\overline{t} $$ t t ¯ pair with mass above 1.4 TeV, exploiting a dedicated multivariate technique with jet substructure to identify hadronically decaying top quarks using large-radius jets and evaluating the background expectation from data. No significant deviation from the background prediction is observed. Limits are set on the production cross-section times branching fraction for the new Z′ boson in a topcolor-assisted-technicolor model. The Z′ boson masses below 3.9 and 4.7 TeV are excluded at 95% confidence level for the decay widths of 1% and 3%, respectively.

Download Full-text

Background subtraction for fluorescence EXAFS data of a very dilute dopantZinZ + 1 host

Journal of Synchrotron Radiation ◽

10.1107/s0909049511012799 ◽

2011 ◽

Vol 18 (4) ◽

pp. 679-680 ◽

Cited By ~ 4

Author(s):

Scott Medling ◽

Frank Bridges

Keyword(s):

Background Subtraction ◽

Raman Peak ◽

Exafs Data ◽

X Ray ◽

Fluorescence Peak ◽

Large Background

When conducting EXAFS at the CuK-edge for ZnS:Cu with very low Cu concentration (<0.04% Cu), a large background was present that increased with energy. This background arises from a Zn X-ray Raman peak, which moves through the Cu fluorescence window, plus the tail of the Zn fluorescence peak. This large background distorts the EXAFS and must be removed separately before reducing the data. A simple means to remove this background is described.

Download Full-text