002764782 001__ 2764782
002764782 005__ 20250510042814.0
002764782 0248_ $$aoai:cds.cern.ch:2764782$$pcerncds:FULLTEXT$$pcerncds:CERN:FULLTEXT$$pcerncds:CERN
002764782 0247_ $$2DOI$$9EDP Sciences$$a10.1051/epjconf/202125103070
002764782 037__ $$9arXiv$$aarXiv:2104.07692$$cquant-ph
002764782 035__ $$9arXiv$$aoai:arXiv.org:2104.07692
002764782 035__ $$9Inspire$$aoai:inspirehep.net:1858884$$d2025-05-09T17:08:16Z$$h2025-05-10T02:07:56Z$$mmarcxml$$ttrue$$uhttps://inspirehep.net/api/oai2d
002764782 035__ $$9Inspire$$a1858884
002764782 041__ $$aeng
002764782 100__ $$aBelis, [email protected]$$uETH, Zurich (main)$$vInstitute of Particle Physics and Astrophysics, ETH Zürich, Zürich, Switzerland
002764782 245__ $$9EDP Sciences$$aHiggs analysis with quantum classifiers
002764782 269__ $$c2021-04-15
002764782 260__ $$c2021
002764782 300__ $$a12 p
002764782 500__ $$9arXiv$$aSubmitted to the 25th International Conference on Computing in High-Energy and Nuclear Physics (vCHEP2021)
002764782 520__ $$9EDP Sciences$$aWe have developed two quantum classifier models for the ttH classification problem, both of which fall into the category of hybrid quantumclassical algorithms for Noisy Intermediate Scale Quantum devices (NISQ). Our results, along with other studies, serve as a proof of concept that Quantum Machine Learning (QML) methods can have similar or better performance, in specific cases of low number of training samples, with respect to conventional ML methods even with a limited number of qubits available in current hardware. To utilise algorithms with a low number of qubits — to accommodate for limitations in both simulation hardware and real quantum hardware — we investigated different feature reduction methods. Their impact on the performance of both the classical and quantum models was assessed. We addressed different implementations of two QML models, representative of the two main approaches to supervised quantum machine learning today: a Quantum Support Vector Machine (QSVM), a kernel-based method, and a Variational Quantum Circuit (VQC), a variational approach.
002764782 520__ $$9arXiv$$aWe have developed two quantum classifier models for the $t\bar{t}H(b\bar{b})$ classification problem, both of which fall into the category of hybrid quantum-classical algorithms for Noisy Intermediate Scale Quantum devices (NISQ). Our results, along with other studies, serve as a proof of concept that Quantum Machine Learning (QML) methods can have similar or better performance, in specific cases of low number of training samples, with respect to conventional ML methods even with a limited number of qubits available in current hardware. To utilise algorithms with a low number of qubits -- to accommodate for limitations in both simulation hardware and real quantum hardware -- we investigated different feature reduction methods. Their impact on the performance of both the classical and quantum models was assessed. We addressed different implementations of two QML models, representative of the two main approaches to supervised quantum machine learning today: a Quantum Support Vector Machine (QSVM), a kernel-based method, and a Variational Quantum Circuit (VQC), a variational approach.
002764782 540__ $$3preprint$$aCC-BY-4.0$$uhttp://creativecommons.org/licenses/by/4.0/
002764782 540__ $$3publication$$aCC-BY-4.0$$bEDP Sciences$$uhttps://creativecommons.org/licenses/by/4.0/
002764782 65017 $$2arXiv$$aphysics.data-an
002764782 65017 $$2SzGeCERN$$aOther Fields of Physics
002764782 65017 $$2arXiv$$ahep-ex
002764782 65017 $$2SzGeCERN$$aParticle Physics - Experiment
002764782 65017 $$2arXiv$$acs.LG
002764782 65017 $$2SzGeCERN$$aComputing and Computers
002764782 65017 $$2arXiv$$aquant-ph
002764782 65017 $$2SzGeCERN$$aGeneral Theoretical Physics
002764782 690C_ $$aCERN
002764782 690C_ $$aARTICLE
002764782 700__ $$aGonzález-Castillo, Samuel$$uOviedo U.$$vFaculty of Sciences, University of Oviedo, Oviedo, Spain
002764782 700__ $$aReissel, Christina$$uETH, Zurich (main)$$vInstitute of Particle Physics and Astrophysics, ETH Zürich, Zürich, Switzerland
002764782 700__ $$aVallecorsa, Sofia$$uCERN$$vCERN, 1, Esplanade des Particules, Geneva, CH 1211
002764782 700__ $$aCombarro, Elías F.$$uOviedo U.$$vDepartment of Computer Science, University of Oviedo, Oviedo, Spain
002764782 700__ $$aDissertori, Günther$$uETH, Zurich (main)$$vInstitute of Particle Physics and Astrophysics, ETH Zürich, Zürich, Switzerland
002764782 700__ $$aReiter, Florentin$$uZurich, ETH-CSCS/SCSC$$vInstitute for Quantum Electronics, ETH Zürich, Zürich, Switzerland
002764782 773__ $$c03070$$pEPJ Web Conf.$$v251$$wC21-05-17.1$$y2021
002764782 8564_ $$82289830$$s23582$$uhttp://cds.cern.ch/record/2764782/files/hist-tf0.png$$y00002 : TensorFlow Auto-Encoder
002764782 8564_ $$82289831$$s26612$$uhttp://cds.cern.ch/record/2764782/files/Classical_HEPApproaches.png$$y00003 The computed ROCs for the BDT and DNN using the test data set. The test data set was split into 5 subsets and for each one the AUC was computed. The mean and uncertainty of the AUCs is depicted in the legend. The models were trained both using the full set of of input features (67) and the reduced set of features (16), constructed from the latent space of one of the two developed Autoencoders (see Sec. \ref{sec:autoencoder}).
002764782 8564_ $$82289832$$s230433$$uhttp://cds.cern.ch/record/2764782/files/zzfm.png$$y00008 Scheme for the feature map used in the VQC implementation, dependent on data feature vectors $\vec{x}$.
002764782 8564_ $$82289833$$s385848$$uhttp://cds.cern.ch/record/2764782/files/2local.png$$y00007 Variational form used in the VQC implementation, dependent on the parameters $\vec{\theta}$.
002764782 8564_ $$82289834$$s53365$$uhttp://cds.cern.ch/record/2764782/files/roc_ae_latent_space.png$$y00009 : Models trained on the AE latent space features (16).
002764782 8564_ $$82289835$$s44889$$uhttp://cds.cern.ch/record/2764782/files/roc_input.png$$y00010 : Models trained on the original input features (67), discarding the 3 least informative ones (64).
002764782 8564_ $$82289836$$s9138$$uhttp://cds.cern.ch/record/2764782/files/Feynman.png$$y00000 Example of Leading Order (LO) Feynman diagram of the signal process in red and the dominant background process in green. The Higgs Boson is produced in association with $t\bar{t}$ via gluon fusion and it decays to $b\bar{b}$. The channel is semi-leptonic as only one of the $W$ bosons decays into leptons.
002764782 8564_ $$82289837$$s45001$$uhttp://cds.cern.ch/record/2764782/files/roc_feature_selection16.png$$y00011 : Models trained on 16 selected features of the input space according to their individual AUC values.
002764782 8564_ $$82289838$$s22462$$uhttp://cds.cern.ch/record/2764782/files/hist-ptReco.png$$y00001 : PyTorch Auto-Encoder
002764782 8564_ $$82289839$$s49278$$uhttp://cds.cern.ch/record/2764782/files/vqc_circuit.png$$y00006 Quantum Circuit for the VQC. $U_\text{enc}$ encodes the data vector $\vec{x}$ into the quantum circuit, then $\ell$ layers of parametrized circuits $G$ and entanglement circuits $U_\text{ent}$ are used. The trainable parameters are $\vec{\theta} = (\vec{\theta}_1,\dots,\vec{\theta}_\ell)$. $\mathcal{O}_{\vec{x},\vec{\theta}}$ is the observable whose expectation value we sample with the quantum device.
002764782 8564_ $$82289840$$s1095279$$uhttp://cds.cern.ch/record/2764782/files/2104.07692.pdf$$yFulltext
002764782 8564_ $$82289841$$s10681$$uhttp://cds.cern.ch/record/2764782/files/u2_reuploading.png$$y00005 Data encoding circuit serving as feature map for the 8-qubit QSVM implementation. The circuit includes generic unitary 1-qubit gates that depend on the elements (16) of the data features vector $\vec{x}$, and 2-qubit Control-X (CNOT) gates accomplishing entanglement.
002764782 8564_ $$82289842$$s53255$$uhttp://cds.cern.ch/record/2764782/files/qsvm_circuit.png$$y00004 Quantum Circuit for the QSVM, where $\vec{x}_i$ and $\vec{x}_j$ are feature vectors of a given pair of data points, $i$ and $j$. The circuit constructs the kernel matrix elements, $K_{ij}$, by sampling the probability of measuring $|0\rangle \coloneqq |0\rangle^{\otimes\, n^\text{qubits}}$, making the kernel (inner product) quantum.
002764782 8564_ $$82318173$$s872203$$uhttp://cds.cern.ch/record/2764782/files/document.pdf$$yFulltext
002764782 960__ $$a13
002764782 962__ $$b2750676$$k03070$$nonline20210517
002764782 980__ $$aConferencePaper
002764782 980__ $$aARTICLE