An Efficient Lorentz Equivariant Graph Neural Network for Jet Tagging

Gong, Shiqi; Meng, Qi; Zhang, Jue; Qu, Huilin; Li, Congqiao; Qian, Sitian; Du, Weitao; Ma, Zhi-Ming; Liu, Tie-Yan

doi:10.48550/arxiv.2201.08187

Cited by 7 publications

(10 citation statements)

References 46 publications

(59 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note added: While this paper was being finalized, we became aware of Ref. [50], which had a similar objective. In contrast to that paper, we use transformers to implement a partial Lorentz covariance -one that is most relevant for hadron colliders.…”

Section: Discussionmentioning

confidence: 99%

A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer

Qiu¹,

Han²,

Ju³

et al. 2022

Preprint

View full text Add to dashboard Cite

Precise reconstruction of top quark properties is a challenging task at the Large Hadron Collider due to combinatorial backgrounds and missing information. We introduce a physics-informed neural network architecture called the Covariant Particle Transformer (CPT) for directly predicting the top quark kinematic properties from reconstructed final state objects. This approach is permutation invariant and partially Lorentz covariant and can account for a variable number of input objects. In contrast to previous machine learning-based reconstruction methods, CPT is able to predict top quark four-momenta regardless of the jet multiplicity in the event. Using simulations, we show that the CPT performs favorably compared with other machine learning top quark reconstruction approaches.

show abstract

Section: Discussionmentioning

confidence: 99%

A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer

Qiu¹,

Han²,

Ju³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…For a given symmetry, one can construct machine learning methods that are invariant or covariant (in machine learning, this is called equivariant) under the action of that symmetry. For example, recent proposals have shown how to construct Lorentz covariant neural networks [42][43][44]. Symmetries can also be used to build a learned representation of a sample [45].…”

Section: Identifying Asymmetries With Neural Networkmentioning

confidence: 99%

Data-directed search for new physics based on symmetries of the SM

et al. 2022

View full text Add to dashboard Cite

We propose exploiting symmetries (exact or approximate) of the Standard Model (SM) to search for physics Beyond the Standard Model (BSM) using the data-directed paradigm (DDP). Symmetries are very powerful because they provide two samples that can be compared without requiring simulation. Focusing on the data, exclusive selections which exhibit significant asymmetry can be identified efficiently and marked for further study. Using a simple and generic test statistic which compares two matrices already provides good sensitivity, only slightly worse than that of the profile likelihood ratio test statistic which relies on the exact knowledge of the signal shape. This can be exploited for rapidly scanning large portions of the measured data, in an attempt to identify regions of interest. We also demonstrate that weakly supervised Neural Networks could be used for this purpose as well.

show abstract

“…Some efforts have also turned towards constructing more expressive features via so-called Energy Flow Polynomials [12,13], or novel QCD inspired features [14,15]. Currently, the most accurate machine learning approaches on jet tagging benchmarks are Lorentz group equivariant message passing networks (LE-MPNNs) [7,16,17]. This approach exploits the fact that the center of mass is accessible through Lorentz boosts and rotations.…”

Section: Center Of Mass Framementioning

confidence: 99%

“…Expanding each f ν in terms of the tensor-product basis φ v1 ⊗ • • • ⊗ φ vν and reorganising the summation (see the SI in V for the details) results exactly in (6) with the ν-correlation features A v arising exactly from the expansion of f ν . Thus, we can alternatively interpret (6) as an efficient linear parametrization of the many-body expansion (7) and the ν-correlation features as natural basis functions for the ν-body term.…”

Section: B Interpretationmentioning

confidence: 99%

See 1 more Smart Citation

BIP: Boost Invariant Polynomials for Efficient Jet Tagging

Munoz¹,

Batatia²,

Ortner³

2022

Preprint

View full text Add to dashboard Cite

Deep Learning approaches are becoming the go-to methods for data analysis in High Energy Physics (HEP). Nonetheless, most physics-inspired modern architectures are computationally inefficient and lack interpretability. This is especially the case with jet tagging algorithms, where computational efficiency is crucial considering the large amounts of data produced by modern particle detectors. In this work, we present a novel, versatile and transparent framework for jet representation; invariant to Lorentz group boosts, which achieves high accuracy on jet tagging benchmarks while being orders of magnitudes faster to train and evaluate than other modern approaches for both supervised and unsupervised schemes.

show abstract

An Efficient Lorentz Equivariant Graph Neural Network for Jet Tagging

Cited by 7 publications

References 46 publications

A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer

A Holistic Approach to Predicting Top Quark Kinematic Properties with the Covariant Particle Transformer

Data-directed search for new physics based on symmetries of the SM

BIP: Boost Invariant Polynomials for Efficient Jet Tagging

Contact Info

Product

Resources

About