Search

We carry out a Bayesian analysis of dark matter (DM) direct detection data to determine particle model parameters using the Truncated Marginal Neural Ratio Estimation (TMNRE) machine learning technique. TMNRE avoids an explicit calculation of the likelihood, which instead is estimated from simulated data, unlike in traditional Markov Chain Monte Carlo (MCMC) algorithms. This considerably speeds up, by several orders of magnitude, the computation of the posterior distributions, which allows to perform the Bayesian analysis of an otherwise computationally prohibitive number of benchmark points. In this article we demonstrate that, in the TMNRE framework, it is possible to include, combine, and remove different datasets in a modular fashion, which is fast and simple as there is no need to re-train the machine learning algorithm or to define a combined likelihood. In order to assess the performance of this method, we consider the case of WIMP DM with spin-dependent and independent interactions with protons and neutrons in a xenon experiment. After validating our results with MCMC, we employ the TMNRE procedure to determine the regions where the DM parameters can be reconstructed. Finally, we present CADDENA, a Python package that implements the modular Bayesian analysis of direct detection experiments described in this work., Comment: 25 pages, 7 figures, 2 tables

The detection of Dark Matter (DM) remains a significant challenge in particle physics. This study exploits advanced machine learning models to improve detection capabilities of liquid xenon time projection chamber experiments, utilizing state-of-the-art transformers alongside traditional methods like Multilayer Perceptrons and Convolutional Neural Networks. We evaluate various data representations and find that simplified feature representations, particularly corrected S1 and S2 signals, retain critical information for classification. Our results show that while transformers offer promising performance, simpler models like XGBoost can achieve comparable results with optimal data representations. We also derive exclusion limits in the cross-section versus DM mass parameter space, showing minimal differences between XGBoost and the best performing deep learning models. The comparative analysis of different machine learning approaches provides a valuable reference for future experiments by guiding the choice of models and data representations to maximize detection capabilities., Comment: 26 pages, 7 figures, 2 tables

In complex industrial and chemical process control rooms, effective decision-making is crucial for safety and efficiency. The experiments in this paper evaluate the impact and applications of an AI-based decision support system integrated into an improved human-machine interface, using dynamic influence diagrams, a hidden Markov model, and deep reinforcement learning. The enhanced support system aims to reduce operator workload, improve situational awareness, and provide different intervention strategies to the operator adapted to the current state of both the system and human performance. Such a system can be particularly useful in cases of information overload when many alarms and inputs are presented all within the same time window, or for junior operators during training. A comprehensive cross-data analysis was conducted, involving 47 participants and a diverse range of data sources such as smartwatch metrics, eye-tracking data, process logs, and responses from questionnaires. The results indicate interesting insights regarding the effectiveness of the approach in aiding decision-making, decreasing perceived workload, and increasing situational awareness for the scenarios considered. Additionally, the results provide valuable insights to compare differences between styles of information gathering when using the system by individual participants. These findings are particularly relevant when predicting the overall performance of the individual participant and their capacity to successfully handle a plant upset and the alarms connected to it using process and human-machine interaction logs in real-time. These predictions enable the development of more effective intervention strategies.

We study the impact of machine-learning algorithms on LHC searches for leptoquarks in final states with hadronically decaying tau leptons, multiple $b$-jets, and large missing transverse momentum. Pair production of scalar leptoquarks with decays only into third-generation leptons and quarks is assumed. Thanks to the use of supervised learning tools with unbinned methods to handle the high-dimensional final states, we consider simple selection cuts which would possibly translate into an improvement in the exclusion limits at the 95$\%$ confidence level for leptoquark masses with different values of their branching fraction into charged leptons. In particular, for intermediate branching fractions, we expect that the exclusion limits for leptoquark masses extend to $\sim$1.3 TeV. As a novelty in the implemented unbinned analysis, we include a simplified estimation of some systematic uncertainties with the aim of studying their possible impact on the stability of the results. Finally, we also present the projected sensitivity within this framework at 14 TeV for 300 and 3000 fb$^{-1}$ that extends the upper limits to $\sim$1.6 and $\sim$1.8 TeV, respectively., Comment: 25 pages, 7 figures, 2 tables (version published in PRD). Unbinned method code with approach to systematic uncertainty inclusion available from https://github.com/AndresDanielPerez/

Machine-Learned Likelihoods (MLL) combines machine-learning classification techniques with likelihood-based inference tests to estimate the experimental sensitivity of high-dimensional data sets. We extend the MLL method by including Kernel Density Estimators (KDE) to avoid binning the classifier output to extract the resulting one-dimensional signal and background probability density functions. We first test our method on toy models generated with multivariate Gaussian distributions, where the true probability distribution functions are known. Later, we apply the method to two cases of interest at the LHC: a search for exotic Higgs bosons, and a $Z'$ boson decaying into lepton pairs. In contrast to physical-based quantities, the typical fluctuations of the ML outputs give non-smooth probability distributions for pure-signal and pure-background samples. The non-smoothness is propagated into the density estimation due to the good performance and flexibility of the KDE method. We study its impact on the final significance computation, and we compare the results using the average of several independent ML output realizations, which allows us to obtain smoother distributions. We conclude that the significance estimation turns out to be not sensible to this issue., Comment: 24 pages, 8 figures, 3 tables, 1 appendix (version published in EPJC). MLL+KDE code available from https://github.com/AndresDanielPerez/2211.04806-ML-Likelihood-with-KDE

We investigate the possibility that right-handed (RH) sneutrinos and gravitinos can coexist and explain the dark matter (DM) problem. We compare extensions of the minimal supersymmetric standard model (MSSM) and the next-to-MSSM (NMSSM) adding RH neutrinos superfields, with special emphasis on the latter. If the gravitino is the lightest supersymmetric particle (LSP) and the RH sneutrino the next-to-LSP (NLSP), the heavier particle decays to the former plus left-handed (LH) neutrinos through the mixing between the scalar partners of the LH and RH neutrinos. However, the interaction is suppressed by the Planck mass, and if the LH-RH sneutrino mixing parameter is small, $\ll O(10^{-2})$, a long-lived RH sneutrino NLSP is possible even surpassing the age of the Universe. As a byproduct, the NLSP to LSP decay produces monochromatic neutrinos in the ballpark of current and planned neutrino telescopes like Super-Kamiokande, IceCube and Antares that we use to set constraints and show prospects of detection. In the NMSSM+RHN, assuming a gluino mass parameter $M_3 = 3$ TeV we found the following lower limits for the gravitino mass $m_{3/2} \gtrsim 1-600$ GeV and the reheating temperature $T_R \gtrsim 10^5 - 3 \times 10^7$ GeV, for $m_{\tilde{\nu}_R} \sim 10-800$ GeV. If we take $M_3=10$ TeV, then the limits on $T_R$ are relaxed by one order of magnitude., Comment: Discussions improved and expanded, version published in JCAP, 35 pages, 7 figures, 3 tables

Machine-learning techniques have become fundamental in high-energy physics and, for new physics searches, it is crucial to know their performance in terms of experimental sensitivity, understood as the statistical significance of the signal-plus-background hypothesis over the background-only one. We present here a simple method that combines the power of current machine-learning techniques to face high-dimensional data with the likelihood-based inference tests used in traditional analyses, which allows us to estimate the sensitivity for both discovery and exclusion limits through a single parameter of interest, the signal strength. Based on supervised learning techniques, it can perform well also with high-dimensional data, when traditional techniques cannot. We apply the method to a toy model first, so we can explore its potential, and then to a LHC study of new physics particles in dijet final states. Considering as the optimal statistical significance the one we would obtain if the true generative functions were known, we show that our method provides a better approximation than the usual naive counting experimental results., Comment: 24 pages, 8 figures; matches version published in Eur. Phys. J. C

In this paper we investigate the $(g-2)_\mu$ discrepancy in the context of the R-parity conserving next-to-minimal supersymmetric Standard Model plus right-handed neutrinos superfields. The model has the ability to reproduce neutrino physics data and includes the interesting possibility to have the right-handed sneutrino as the lightest supersymmetric particle and a viable dark matter candidate. Since right-handed sneutrinos are singlets, no new contributions for $\delta a_{\mu}$ with respect to the MSSM and NMSSM are present. However, the possibility to have the right-handed sneutrino as the lightest supersymmetric particle opens new ways to escape Large Hadron Collider and direct detection constraints. In particular, we find that dark matter masses within $10 \lesssim m_{\tilde{\nu}_{R}} \lesssim 600$ GeV are fully compatible with current experimental constraints. Remarkably, not only spectra with light sleptons are needed, but we obtain solutions with $m_{\tilde{\mu}} \gtrsim 600$ GeV in the entire dark matter mass range that could be probed by new $(g-2)_\mu$ data in the near future. In addition, dark matter direct detection experiments will be able to explore a sizable portion of the allowed parameter space with $m_{\tilde{\nu}_{R}} \lesssim 300$ GeV, while indirect detection experiments will be able to probe a much smaller fraction within $200 \lesssim m_{\tilde{\nu}_{R}} \lesssim 350$ GeV., Comment: Version published in Nucl. Phys. B. Discussions expanded, appendix B and references added. 29 pages, 9 figures, 2 tables. Analysis of the viability of RH sneutrino as dark matter carried out in arXiv:2102.08986

We study several simplified dark matter (DM) models and their signatures at the LHC using neural networks. We focus on the usual monojet plus missing transverse energy channel, but to train the algorithms we organize the data in 2D histograms instead of event-by-event arrays. This results in a large performance boost to distinguish between standard model (SM) only and SM plus new physics signals. We use the kinematic monojet features as input data which allow us to describe families of models with a single data sample. We found that the neural network performance does not depend on the simulated number of background events if they are presented as a function of $S/\sqrt{B}$, where $S$ and $B$ are the number of signal and background events per histogram, respectively. This provides flexibility to the method, since testing a particular model in that case only requires knowing the new physics monojet cross section. Furthermore, we also discuss the network performance under incorrect assumptions about the true DM nature. Finally, we propose multimodel classifiers to search and identify new signals in a more general way, for the next LHC run., Comment: 52 pages, 29 figures, 3 tables. v2: Added a subsection to briefly discuss and compare event-by-event combination versus histogram or ensemble classifiers. v3: minor changes

R-parity conserving supersymmetric models with right-handed (RH) neutrinos are very appealing since they could naturally explain neutrino physics and also provide a good dark matter (DM) candidate such as the lightest supersymmetric particle (LSP). In this work we consider the next-to-minimal supersymmetric standard model (NMSSM) plus RH neutrino superfields, with effective Majorana masses dynamically generated at the electroweak scale (EW). We perform a scan of the relevant parameter space and study both possible DM candidates: RH sneutrino and neutralino. Especially for the case of RH sneutrino DM we analyse the intimate relation between both candidates to obtain the correct amount of relic density. Besides the well-known resonances, annihilations through scalar quartic couplings and coannihilation mechanisms with all kind of neutralinos, are crucial. Finally, we present the impact of current and future direct and indirect detection experiments on both DM candidates., Comment: Version published in JCAP, 40 pages, 8 figures, 6 tables

To elucidate the composition of dark matter (DM) is one of the most important open questions in particle and astroparticle phenomenology. Within the framework of minimal supersymmetric extensions of the standard model of fundamental particles, the `mu-from-nu supersymmetric standard model', $\mu\nu$SSM, solves the $\mu$-problem and reproduces neutrino data, only adding couplings involving right-handed neutrinos. In this context, the supersymmetric partner of the graviton, known as gravitino, becomes a natural DM candidate. Working in the mentioned model, in this thesis we have analyzed the detection of gamma-ray signals arising from decaying gravitino DM considering experiments such as $\textit{Fermi}$-LAT. For this purpose all gravitino decays have been taken into account, complementing previous works. The possibility of an upcoming new generation of gamma-ray detectors has motivated the inclusion of axino, the supersymmetric partner of the axion, as a DM candidate. In this thesis, axino DM has been analyzed in the context of the $\mu\nu$SSM for the first time. Finally, scenarios with multicomponent DM have been studied, focusing on gravitinos and axinos coexisting as DM constituents. This novel scenario includes two very different cases with distinctive features: axino lighter than gravitino and vice versa. Considering the sensitivity of future MeV-GeV gamma-ray telescopes, axino and gravitino prospects of detection were analyzed in the mixed scenarios mentioned previously. In some parameter space regions, both candidates can produce a signal, therefore a double-line feature arises as a smoking gun that would uncover the DM composition., Comment: PhD thesis, 187 pages, in Spanish (Abstract in English). Based on arXiv:1608.08640, arXiv:1911.03191, and arXiv:1911.08550

In $R$-parity violating supersymmetry, the gravitino as the lightest supersymmetric particle (LSP) is a good candidate for dark matter, with the interesting characteristic to be detectable through $\gamma$-ray telescopes. We extend this analysis considering an axino next-to-LSP (NLSP) as a coexisting dark matter particle contributing with a detectable signal in the $\gamma$-ray spectrum. The analysis is carried out in the framework of the $\mu\nu$SSM, which solves the $\mu$ problem reproducing simultaneously neutrino data only with the addition of right-handed neutrinos. We find that important regions of the parameter space can be tested by future MeV-GeV $\gamma$-ray telescopes through the line signal coming from the decay of the axino NLSP into photon-neutrino. In a special region, a double-line signal from axino NLSP and gravitino LSP is possible with both contributions detectable., Comment: Version published in Astroparticle Physics. Discussions expanded, appendix and references added, 23 pages, 4 figures. The analysis of the opposite case (axino LSP with gravitino NLSP) was carried out in arXiv:1911.03191

Axino and gravitino are promising candidates to solve the dark matter (DM) problem in the framework of supersymmetry. In this work, we assume that the axino is the lightest supersymmetric particle (LSP), and therefore contributes to DM. In the case of R-parity violating models, the axino can decay into a neutrino-photon pair with a lifetime much longer than the age of the Universe, yielding a potentially detectable signal. Interestingly, a gravitino next-to-LSP (NLSP) can live enough as to contribute to the relic density. We study both scenarios, only axino LSP as DM, and axino LSP with gravitino NLSP as DM. We carry out the analysis in the context of the $\mu\nu$SSM, which solves the $\mu$ problem and reproduces neutrino data, only adding couplings involving right-handed neutrinos. In particular, we perform a complete analysis of the relevant parameter space of the model considering constraints from neutrino physics, cosmological observations, and $\gamma$-ray detection. We find that the axino or the gravitino can produce a signal detectable by future MeV-GeV $\gamma$-ray telescopes. In addition, in a parameter region where we get a well-tempered mixture of both particles, a double-line signal arises as a smoking gun., Comment: Version published in JCAP, discussions expanded, references added, 27 pages, 6 figures

In this paper, we investigate how to learn rich and robust feature representations for audio classification from visual data and acoustic images, a novel audio data modality. Former models learn audio representations from raw signals or spectral data acquired by a single microphone, with remarkable results in classification and retrieval. However, such representations are not so robust towards variable environmental sound conditions. We tackle this drawback by exploiting a new multimodal labeled action recognition dataset acquired by a hybrid audio-visual sensor that provides RGB video, raw audio signals, and spatialized acoustic data, also known as acoustic images, where the visual and acoustic images are aligned in space and synchronized in time. Using this richer information, we train audio deep learning models in a teacher-student fashion. In particular, we distill knowledge into audio networks from both visual and acoustic image teachers. Our experiments suggest that the learned representations are more powerful and have better generalization capabilities than the features learned from models trained using just single-microphone audio data., Comment: Accepted at WACV 2020; supplementary material at page 11; code available at https://github.com/afperezm/acoustic-images-distillation