Interpretability of an Interaction Network for identifying $H \rightarrow b\bar{b}$ jets

Multivariate techniques and machine learning models have found numerous applications in High Energy Physics (HEP) research over many years. In recent times, AI models based on deep neural networks are becoming increasingly popular for many of these applications. However, neural networks are regarded as black boxes- because of their high degree of complexity it is often quite difficult to quantitatively explain the output of a neural network by establishing a tractable input-output relationship and information propagation through the deep network layers. As explainable AI (xAI) methods are becoming more popular in recent years, we explore interpretability of AI models by examining an Interaction Network (IN) model designed to identify boosted

H \to b\bar{b}

jets amid QCD background. We explore different quantitative methods to demonstrate how the classifier network makes its decision based on the inputs and how this information can be harnessed to reoptimize the model- making it simpler yet equally effective. We additionally illustrate the activity of hidden layers within the IN model as Neural Activation Pattern (NAP) diagrams. Our experiments suggest NAP diagrams reveal important information about how information is conveyed across the hidden layers of deep model. These insights can be useful to effective model reoptimization and hyperparameter tuning.

Note:

Contribution to Proceedings of 41st International Conference on High Energy Physics - ICHEP2022

Higgs particle: hadronic decay
bottom: pair production
Higgs particle: boosted particle
quantum chromodynamics: background
network
neural network
propagation
artificial intelligence
machine learning
data analysis method

References(7)

Figures(8)

[1]

Explanation in artificial intelligence: Insights from the social sciences

T. Miller

- Artificial Intel. 267 (2019) 1
•
- https://doi.org/https://doi.org/10.1016/j.artint.2018.07.007

[2]

XAI—Explainable artificial intelligence, Science Robotics 4 eaay7120

D. Gunning
,
M. Stefik
,
J. Choi
,
T. Miller
,
S. Stumpf

et al.

[3]

A survey of interpretability of machine learning in accelerator-based high energy physics, in

D. Turvill
,
L. Barnby
,
B. Yuan
,
A. Zahir

•
DOI:
- 10.1109/BDCAT50828.2020.00025

[4]

Interaction networks for the identification of boosted $H \rightarrow b\overline{b}$ decays

et al.

- Phys.Rev.D 102 (2020) 1, 012010
•
e-Print:
- 1909.12285
•
DOI:
- 10.1103/PhysRevD.102.012010

[5]

Layer-wise relevance propagation for deep neural network architectures, in Information science and applications (ICISA), pp. 913-922

A. Binder
,
S. Bach
,
G. Montavon
,
K.-R. Müller
,
W. Samek

[6]

Layer-wise relevance propagation: an overview, Explainable AI: interpreting, explaining and visualizing deep learning 193

G. Montavon
,
A. Binder
,
S. Lapuschkin
,
W. Samek
,
K.-R. Müller

[7]

A detailed study of interpretability of deep neural network based top taggers

Ayush Khot
(
- NCSA, Urbana and
- Illinois U., Urbana
)
,
Mark S. Neubauer
(
- NCSA, Urbana and
- Illinois U., Urbana
)
,
Avik Roy
(
- NCSA, Urbana and
- Illinois U., Urbana
)

- Mach.Learn.Sci.Tech. 4 (2023) 3, 035003
•
e-Print:
- 2210.04371
•
DOI:
- 10.1088/2632-2153/ace0a1

Interpretability of an Interaction Network for identifying H→bbˉH \rightarrow b\bar{b}H→bbˉ jets

Citations per year

Interaction networks for the identification of boosted H→bb‾H \rightarrow b\overline{b}H→bb decays

A detailed study of interpretability of deep neural network based top taggers

Interpretability of an Interaction Network for identifying $H \rightarrow b\bar{b}$ jets

Interaction networks for the identification of boosted $H \rightarrow b\overline{b}$ decays