We compare the performance of a convolutional neural network (CNN) trained on jet images with dense neural networks (DNNs) trained on n-subjettiness variables to study the distinguishing power of these two separate techniques applied to top quark decays. We find that they perform almost identically and are highly correlated once jet mass information is included, which suggests they are accessing the same underlying information which can be intuitively understood as being contained in 4-, 5-, 6-, and 8-body kinematic phase spaces depending on the sample. This suggests both of these methods are highly useful for heavy object tagging and provides a tentative answer to the question of what the image network is actually learning.
Cited by 4
Sven Bollweg et al., Deep-learning jets with uncertainties and more
SciPost Phys. 8, 006 (2020) [Crossref]
Sascha Diefenbacher et al., CapsNets continuing the convolutional quest
SciPost Phys. 8, 023 (2020) [Crossref]
M. Aaboud et al., Performance of top-quark and
-boson tagging with ATLAS in Run 2 of the LHC
Eur. Phys. J. C 79, 375 (2019) [Crossref]
Christian Bierlich et al., Robust Independent Validation of Experiment and Theory: Rivet version 3
SciPost Phys. 8, 026 (2020) [Crossref]
Ontology / TopicsSee full Ontology or Topics database.
Authors / Affiliations: mappings to Contributors and OrganizationsSee all Organizations.
- 1 Université catholique de Louvain [UCL]
- 2 Laboratoire de Physique Théorique et Hautes Energies / Laboratory of Theoretical and High Energy Physics [LPTHE]
- 3 Nationaal instituut voor Subatomaire Fysica / National Institute for Subatomic Physics [NIKHEF]
- 4 King's College London [KCL]
- European Research Council [ERC]
- Fonds De La Recherche Scientifique - FNRS (FNRS) (through Organization: Fonds National de la Recherche Scientifique [FNRS])
- Nederlandse Organisatie voor Wetenschappelijk Onderzoek / Netherlands Organisation for Scientific Research [NWO]
- Science and Technology Facilities Council [STFC]