Jet Flavor Classification in High-Energy Physics with Deep Neural Networks

Classification of jets as originating from light-flavor or heavy-flavor quarks is an important task for inferring the nature of particles produced in high-energy collisions. The large and variable dimensionality of the data provided by the tracking detectors makes this task difficult. The current state-of-the-art tools require expert data reduction to convert the data into a fixed low-dimensional form that can be effectively managed by shallow classifiers. We study the application of deep networks to this task, attempting classification at several levels of data, starting from a raw list of tracks. We find that the highest-level lowest-dimensionality expert information sacrifices information needed for classification, that the performance of current state-of-the-art taggers can be matched or slightly exceeded by deep-network-based taggers using only track and vertex information, that classification using only lowest-level highest-dimensionality tracking information remains a difficult task for deep networks, and that adding lower-level track and vertex information to the classifiers provides a significant boost in performance compared to the state of the art.

Note:

12 pages, submitted to PRD

jet: flavor
new physics: search for
high energy behavior
neural network: performance
tracking detector
heavy quark: jet
jet: charm
jet: production

References(42)

Figures(47)

[1]

Georges A search for top squarks with Rparity-violating decays to all-hadronic final states with the ATLAS detector in √ s = 8 TeV proton-proton collisions

Aad

[2]

Search for single production of a vector-like quark via a heavy gluon in the $4b$ final state with the ATLAS detector in $pp$ collisions at $\sqrt{s} = 8$ TeV

ATLAS

Collaboration

•

Georges Aad
(
- Marseille, CPPM
)

et al.

- Phys.Lett.B 758 (2016) 249-268
•
e-Print:
- 1602.06034
•
DOI:
- 10.1016/j.physletb.2016.04.061

[3]

The RAVE/VERTIGO vertex reconstruction toolkit and framework

- J.Phys.Conf.Ser. 119 (2008) 032037
•
DOI:
- 10.1088/1742-6596/119/3/032037

[4]

Performance of $b$ -Jet Identification in the ATLAS Experiment

ATLAS

Collaboration

•

Georges Aad
(
- Marseille, CPPM
)

et al.

- JINST 11 (2016) 04, P04008
•
e-Print:
- 1512.01094
•
DOI:
- 10.1088/1748-0221/11/04/P04008

[5]

Identification of b-Quark Jets with the CMS Experiment

CMS

Collaboration

•

Serguei Chatrchyan
(
- Yerevan Phys. Inst.
)

et al.

- JINST 8 (2013) P04013
•
e-Print:
- 1211.4462
•
DOI:
- 10.1088/1748-0221/8/04/P04013

[6]

Pierre Baldi, Peter Sadowski, and Daniel Whiteson. Searching for exotic particles in high-energy physics with deep learning. Nature Communications, 5, July

[7]

Enhanced Higgs Boson to $\tau^+\tau^-$ Search with Deep Learning

- Phys.Rev.Lett. 114 (2015) 11, 111801
•
e-Print:
- 1410.3469
•
DOI:
- 10.1103/PhysRevLett.114.111801

[8]

Journal of Machine Learning Research, Workshop and Jet Minimum Pseudorapidity 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 Light-quarkRejection 1 10 2 10 3 10 4 10 Tracks+Vertices+Expert Expert Eff=0.75 Eff=0.60 Jet Minimum Pseudorapidity 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 c-quarkRejection 1 10 2 10

P. Sadowski
,
J. Collado
,
D. Whiteson
,
P. Baldi

- Conf.Proc. 42 (2015) 81-97

[9]

Pierre Baldi, Kevin Bauer, Clara Eng, Peter Sadowski, and Daniel Whiteson. Jet Substructure Classification in High-Energy Physics with Deep Neural Networks

[10]

Pierre Baldi and Yves Chauvin. Hybrid modeling, hmm/nn architectures, and protein applications

- Neural Comput. 8 (1996) 1541-1565

[11]

Learning task-dependent distributed representations by backpropagation through structure International Conference on Neural Networks, page 347352

C. Goller
,
A. Kuchler

[12]

A general framework for adaptive processing of data structures. Trans. Neur. Netw., 9(5):768-786, September

P. Frasconi
,
M. Gori
,
A. Sperduti

[13]

Neural Computation

F.A. Gers
,
J. Schmidhuber
,
F. Cummins

[14]

Pierre Baldi and Gianluca Pollastri. The principled design of large-scale recursive neural network architectures- dag-rnns and the protein structure prediction problem Learn. Res., 4:575-602, December

J. Mach

[15]

Proceedings of the conference on empirical methods in natural language processing, 1631:1642

R. Socher
,
A. Perelygin
,
J.Y. Wu
,
J. Chuang
,
C.D. Manning

et al.

[16]

Exploiting the past and the future in protein secondary structure prediction. Bioinformatics, 15:937946,. [GeV] T Jet p 0 50 100 150 200 250 300 FractionofSelectedEvents -4 10 -3 10 -2 10 -1 10 b-quark Quark, rej=50 (expert) Quark, rej=50 (LL+expert) Quark, rej=10 (expert) Quark, rej=10 (LL+expert) Quark Jet Pseudorapidity -3 -2 -1 0 1 2 3 4

P. Baldi
,
S. Brunak
,
P. Frasconi
,
G. Pollastri
,
G. Soda

[17]

Nncon: improved protein contact map prediction using 2d-recursive neural networks

A.N. Tegge
,
Z. Wang
,
J. Eickholt
,
J. Cheng

- Nucleic Acids Res. 37 (2009) 515518

[18]

Deep architectures for protein contact map prediction. Bioinformatics, 28:24492457

P. Di Lena
,
K. Nagata
,
P. Baldi

[19]

Sspro/accpro 5: Almost perfect prediction of protein secondary structure and relative solvent accessibility using profiles, machine learning, and structural similarity. Bioinformatics, 30:25922597

C.N. Magnan
,
P. Baldi

[20]

Deep architectures and deep learning in chemoinformatics: the prediction of aqueous solubility for drug-like molecules. Journal of chemical information and modeling, 53:15631575

A. Lusci
,
G. Pollastri
,
P. Baldi

[21]

Dougal Maclaurin, Jorge Iparraguirre, Rafael Bombarell, Timothy Hirzel, Alan Aspuru-Guzik, and Ryan P Adams. Convolutional networks on graphs for learning molecular fingerprints. In C. Cortes, N. D. Lawrence, D. D. Lee, M. Sugiyama, and R. Garnett (eds.) Advances in Neural Information Processing Systems 28, pages 2215-2223. Curran Associates, Inc

David K. Duvenaud

[22]

Lin Wu and Pierre Baldi. Learning to play go using recursive neural networks

- Neural Networks 21 (2008) 1392-1400

[23]

395374/,. DataScience@LHC

L. de Oliveira

[24]

DataScience@LHC

P. Baldi

[25]

MadGraph 5 : Going Beyond

Johan Alwall
(
- Fermilab
)
,
Michel Herquet
(
- NIKHEF, Amsterdam
)
,
Fabio Maltoni
(
- Louvain U., CP3
)
,
Olivier Mattelaer
(
- Louvain U., CP3
)
,
Tim Stelzer
(
- Illinois U., Urbana
)

- JHEP 06 (2011) 128
•
e-Print:
- 1106.0522
•
DOI:
- 10.1007/JHEP06(2011)128

1-25 of 42
1
2
25 / page