380 Search Results

As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multiple efforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource andmore » latency estimation, and its corresponding initial dataset of over 680,000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75% percentile within several percent of the synthesized resources on the synthetic test dataset.« less

A search is performed for dark matter produced in association with bottom quarks and a pair of electrons or muons in data collected with the CMS detector at the LHC, corresponding to 138 fb$$^{−1}$$ of integrated luminosity of proton-proton collisions at a center-of-mass energy of 13 TeV. For the first time at the LHC, the associated production of a bottom quark-antiquark pair and a new heavy neutral Higgs boson (H) that subsequently decays into a leptonically decaying Z boson and a pseudoscalar (a) is explored. The latter acts as a dark matter mediator in the context of the two Higgsmore » doublet model plus a pseudoscalar (2HDM+a). Multivariate techniques that target a wide range of mass configurations for the H and a particles are used. The observations are consistent with the expectations from standard model processes. Upper limits at 95% confidence level are set on the product of cross section and branching fraction of the new particles, ranging from 10$$^{−2}$$ pb for an H mass of 400 GeV to 10$$^{−3}$$ pb for an H mass of 2000 GeV. Constraints on the parameter space of a benchmark 2HDM+a model are derived and compared with expectations in the context of cosmological predictions.[graphic not available: see fulltext]« less

A search for the pair production of heavy spin- $1/2$ or spin- $3/2$ resonances ( $t^{*}$ ) in proton-proton collisions at $\sqrt{s}=13\mathrm{TeV}$ is presented. Data collected with the CMS detector at the CERN LHC from 2016 to 2018 corresponding to an integrated luminosity of $138{\mathrm{fb}}^{-1}$ are used. The analysis targets benchmark signal scenarios where one $t^{*}$ decays into a top quark ( $t$ ) and a photon ( $\gamma$ ), and the other into a $t$ quarkmore » and a gluon ( $g$ ), i.e., $pp\to t^{*}\overline{t^{*}}\to tt\gamma g$ . All-hadronic final states from the $t$ pair decay chain are selected using jet substructure techniques. The signal is probed as a function of the $t^{*}$ candidate mass, which is reconstructed using the photon and a top quark candidate jet. No significant deviation from the background-only hypothesis is found. Observed (expected) upper limits on the signal cross section at 95% confidence level are set, excluding masses of spin- $1/2$ $t^{*}$ particles below 930 (930) GeV and spin- $3/2$ $t^{*}$ particles below 1330 (1390) GeV. This analysis marks the first search for heavy resonances in the $t\overline{t}\gamma g$ channel. Exploiting the high-energy photon to reduce the backgrounds, this search achieves sensitivity competitive with $pp\to t^{*}\overline{t^{*}}\to t\overline{t}gg$ searches for spin- $1/2$ $t^{*}$ despite the small expected $t^{*}\to t\gamma$ branching fraction.« less

The ratio between the B$$_\mathrm{c}^+$$$$\to$$ J/$$ψ$$$τ^+ν_τ$$ and B$$_\mathrm{c}^+$$$$\to$$ J/$$ψ$$$μ^+ν_μ$$ branching fractions is measured using a data sample of proton-proton collisions collected by CMS at a center-of-mass energy of 13 TeV in the years 2016$$-$$2018 and corresponding to an integrated luminosity of 138 fb$$^{-1}$$. The J$/ψ$ meson is identified through its J$/ψ$$$$\to$$$$μ^+μ^-$ decay and the tau lepton is reconstructed in the hadronic three-prong final state. The measured ratio of branching fractions in this tau decay mode, $$\mathcal{R}^\text{had}_{\mathrm{J}/ψ}$$ = 1.04 $$_{-0.44}^{+0.50}$$, is combined with the previous analysis based on the $τ^+$$$$\to$$$$μ^+ν_μ\barν_τ$$ leptonic decay channel, leading to $$\mathcal{R}_{\mathrm{J}/ψ}$$ = 0.49 $$\pm$$ 0.26. Asmore » this result is consistent with the standard model prediction of 0.258 $$\pm$$ 0.004, no evidence of lepton flavor universality violation is found.« less

A novel anomaly detection algorithm is presented. The Wasserstein normalized autoencoder (WNAE) is a normalized probabilistic model that minimizes the Wasserstein distance between the learned probability distribution—a Boltzmann distribution where the energy is the reconstruction error of the autoencoder (AE)—and the distribution of the training data. This algorithm has been developed and applied to the identification of semivisible jets—conical sprays of visible standard model (SM) particles and invisible dark matter states—with the CMS experiment at the CERN LHC. Trained on jets of particles from simulated SM processes, the WNAE is shown to learn the probability distribution of the input datamore » in a fully unsupervised fashion, such that it effectively identifies new physics jets as anomalies. The model exhibits stable, convergent training and recovers strong classification performance for a wide range of signals against the selected background process, for which a standard AE fails because of outlier reconstruction. In addition, the model improves upon standard normalized autoencoders while remaining fully agnostic to the signal. The WNAE directly tackles the problem of outlier reconstruction, a common failure mode of autoencoders in anomaly detection tasks.« less

The Higgs boson (H) decay width is determined from the ratio of off- and on-shell production of H $$\to$$ WW $$\to$$ e$νμν$ using proton-proton collision data corresponding to an integrated luminosity of 138 fb$$^{-1}$$ collected at $$\sqrt{s}$$ = 13 TeV by the CMS experiment at the Large Hadron Collider. The off-shell signal strength is measured as $$μ_\text{off-shell}$$ = 1.2 $$^{+0.8}_{-0.7}$$. The Higgs boson total decay width is $$Γ_\text{H}$$ = 3.9 $$^{+2.7}_{-2.2}$$ MeV, in agreement with the standard model prediction. The uncertainty in this result represents a factor of three improvement over the previous CMS result in this decay channel.

Differential top quark pair production cross sections are measured in the dilepton final states e$$^{+}$$e$$^{−}$$, μ$$^{+}$$μ$$^{−}$$, and e$$^{±}$$μ$$^{∓}$$, as a function of kinematic variables of the two-neutrino system: the transverse momentum $$ {p}_{\textrm{T}}^{\nu \nu} $$ of the dineutrino system, the minimum distance in azimuthal angle between $$ {\overrightarrow{p}}_{\textrm{T}}^{\nu \nu} $$ and leptons, and in two dimensions in bins of both observables. The measurements are performed using CERN LHC proton-proton collisions at $$ \sqrt{s}=13 $$ TeV, recorded by the CMS detector between 2016 and 2018, corresponding to an integrated luminosity of 138 fb$$^{−1}$$. The measured cross sections are unfolded to themore » particle level using an unregularized least squares method. Results are compared with predictions by the standard model of particle physics, and found to be in agreement with theoretical calculations as well as Monte Carlo simulations.[graphic not available: see fulltext]« less

The results of a search for the production of two scalar bosons in final states with two photons and two tau leptons are presented. The search considers both nonresonant production of a Higgs boson pair, HH, and resonant production via a new boson X which decays either to HH or to H and a new scalar Y. The analysis uses up to 138 fb$$^{−1}$$ of proton-proton collision data, recorded between 2016 and 2018 by the CMS experiment at the LHC at a center-of-mass energy of 13 TeV. No evidence for signal is found in the data. For the nonresonant production,more » the observed (expected) upper limit at 95% confidence level (CL) on the HH production cross section is set at 930 (740) fb, corresponding to 33 (26) times the standard model prediction. At 95% CL, HH production is observed (expected) to be excluded for values of κ$$_{λ}$$ outside the range between −12 (−9.4) and 17 (15). Observed (expected) upper limits at 95% CL for the X → HH cross section are found to be within 160 to 2200 (200 to 1800) fb, depending on the mass of X. In the X → Y(ττ)H(γγ) search, the observed (expected) upper limits on the product of the production cross section and decay branching fractions vary between 0.059–1.2 fb (0.087–0.68 fb). For the X → Y(γγ)H(ττ) search the observed (expected) upper limits on the product of the production cross section and Y → γγ branching fraction vary between 0.69–15 fb (0.73–8.3 fb) in the low Y mass search, tightening constraints on the next-to-minimal supersymmetric standard model, and between 0.64–10 fb (0.70–7.6 fb) in the high Y mass search.[graphic not available: see fulltext]« less

A measurement of the angular structure of inclusive jets and those containing a prompt D$$^{0}$$ meson in proton-proton collisions at the LHC at a center-of-mass energy of 5.02 TeV is presented. The data corresponding to an integrated luminosity of 301 pb$$^{−1}$$ were collected by the CMS experiment in 2017. Two jet grooming algorithms, late-k$$_{T}$$ and soft drop, are used to study the intrajet radiation pattern using iterative Cambridge-Aachen declustering. The splitting-angle distributions of jets with transverse momentum (p$$_{T}$$) of around 100 GeV, obtained with these two algorithms, show that there is a shift of the distribution for jets containing amore » prompt D$$^{0}$$ meson with respect to inclusive jets. The suppression of emissions at small angles observed in the late-k$$_{T}$$ grooming approach is consistent with the dead-cone effect, whereas the similar suppression for splittings selected with the soft-drop algorithm appears to be induced by gluon splitting to charm quark-antiquark pairs at large angles. The measured distributions are corrected to the particle level and can be used to constrain model predictions for the substructure of high-p$$_{T}$$ charm quark jets.[graphic not available: see fulltext]« less

A search for long-lived particles using final states including a displaced vertex with low-momentum tracks, large missing transverse momentum, and a jet from initial-state radiation is presented. This search uses proton-proton collision data at a center-of-mass energy of 13 TeV collected by the CMS experiment at the CERN LHC in 2017 and 2018, with a total integrated luminosity of 100 fb$$^{−1}$$. This analysis adopts specific supersymmetric (SUSY) coannihilation scenarios as benchmark signal models, characterized by a next-to-lightest SUSY particle (NLSP) with a mass difference of less than 25 GeV relative to the lightest SUSY particle, assumed to be a bino-likemore » neutralino. In the top squark $$ \left(\overset{\sim }{\mathrm{t}}\right) $$ NLSP model, the NLSP is a long-lived $$ \overset{\sim }{\mathrm{t}} $$, while in the bino-wino NLSP scenario, the mass-degenerate NLSPs are a wino-like long-lived neutralino and a short-lived chargino. The search excludes top squarks with masses less than 400–1100 GeV and wino-like neutralinos with masses less than 220–550 GeV, depending on the signal parameters, including the mass difference, mass, and lifetime of the long-lived particle. It sets the most stringent limits to date for the $$ \overset{\sim }{\mathrm{t}} $$ and bino-wino NLSP models.[graphic not available: see fulltext]« less