|
|
ANTARES Collaboration(Albert, A. et al), Alves, S., Calvo, D., Carretero, V., Gozzini, R., Hernandez-Rey, J. J., et al. (2026). Deep learning framework for enhanced neutrino reconstruction of single-line events in the ANTARES telescope. Mach. Learn.-Sci. Technol., 7(3), 035004–27pp.
Abstract: We present the N-fit algorithm designed to improve the reconstruction of neutrino events detected by a single line of the ANTARES underwater telescope, usually associated with low energy neutrino events (similar to 100 GeV). N-Fit is a neural network model that relies on deep learning and combines several advanced techniques in machine learning-deep convolutional layers, mixture density output layers, and transfer learning (TL). This framework divides the reconstruction process into two dedicated branches for each neutrino event topology-tracks and showers-composed of sub-models for spatial estimation-direction and position-and energy inference, which later on are combined for event classification. Regarding the direction of single-line (SL) events, the N-Fit algorithm significantly refines the estimation of the zenithal angle, and delivers reliable azimuthal angle predictions that were previously unattainable with traditional chi 2-fit methods. Improving on energy estimation of SL events is a tall order; N-Fit benefits from TL to efficiently integrate key characteristics, such as the estimation of the closest distance from the event to the detector. N-Fit also takes advantage from TL in event topology classification by freezing convolutional layers of the pretrained branches. Tests on Monte Carlo simulations and data demonstrate a significant reduction in mean and median absolute errors across all reconstructed parameters. The improvements achieved by N-Fit highlight its potential for advancing multimessenger astrophysics and enhancing our ability to probe fundamental physics beyond the Standard Model using SL events from ANTARES data.
|
|
|
Babiano, V., Caballero, L., Calvo, D., Ladarescu, I., Olleros, P., & Domingo-Pardo, C. (2019). gamma-Ray position reconstruction in large monolithic LaCl3(Ce) crystals with SiPM readout. Nucl. Instrum. Methods Phys. Res. A, 931, 1–22.
Abstract: We report on the spatial response characterization of large LaCl3(Ce) monolithic crystals optically coupled to 8 x 8 pixel silicon photomultiplier (SiPM) sensors. A systematic study has been carried out for 511 keV gamma-rays using three different crystal thicknesses of 10 mm, 20 mm and 30 mm, all of them with planar geometry and a base size of 50 x 50 mm(2). In this work we investigate and compare two different approaches for the determination of the main gamma-ray hit location. On one hand, methods based on the fit of an analytical model for the scintillation light distribution provide the best results in terms of linearity and field of view, with spatial resolutions close to similar to 1 mm FWHM. On the other hand, position reconstruction techniques based on neural networks provide similar linearity and field-of-view, becoming the attainable spatial resolution similar to 3 mm FWHM. For the third space coordinate z or depth-of-interaction we have implemented an inverse linear calibration approach based on the cross-section of the measured scintillation-light distribution at a certain height. The detectors characterized in this work are intended for the development of so-called Total Energy Detectors with Compton imaging capability (i-TED), aimed at enhanced sensitivity and selectivity measurements of neutron capture cross sections via the time-of-flight (TOF) technique.
|
|
|
Balibrea-Correa, J., Lerendegui-Marco, J., Babiano-Suarez, V., Caballero, L., Calvo, D., Ladarescu, I., et al. (2021). Machine Learning aided 3D-position reconstruction in large LaCl3 crystals. Nucl. Instrum. Methods Phys. Res. A, 1001, 165249–17pp.
Abstract: We investigate five different models to reconstruct the 3D gamma-ray hit coordinates in five large LaCl3(Ce) monolithic crystals optically coupled to pixelated silicon photomultipliers. These scintillators have a base surface of 50 x 50 mm(2) and five different thicknesses, from 10 mm to 30 mm. Four of these models are analytical prescriptions and one is based on a Convolutional Neural Network. Average resolutions close to 1-2 mm fwhm are obtained in the transverse crystal plane for crystal thicknesses between 10 mm and 20 mm using analytical models. For thicker crystals average resolutions of about 3-5 mm fwhm are obtained. Depth of interaction resolutions between 1 mm and 4 mm are achieved depending on the distance of the interaction point to the photosensor surface. We propose a Machine Learning algorithm to correct for linearity distortions and pin-cushion effects. The latter allows one to keep a large field of view of about 70%-80% of the crystal surface, regardless of crystal thickness. This work is aimed at optimizing the performance of the so-called Total Energy Detector with Compton imaging capability (i-TED) for time-of-flight neutron capture cross-section measurements.
|
|
|
Dilator, J. M., Jaworski, G., Goasduff, A., Gonzalez, V., Gadea, A., Palacz, M., et al. (2025). Reconstruction of pile-up events using a one-dimensional convolutional autoencoder for the NEDA detector array. Nucl. Sci. Tech., 36(2), 32–9pp.
Abstract: Pulse pile-up is a problem in nuclear spectroscopy and nuclear reaction studies that occurs when two pulses overlap and distort each other, degrading the quality of energy and timing information. Different methods have been used for pile-up rejection, both digital and analogue, but some pile-up events may contain pulses of interest and need to be reconstructed. The paper proposes a new method for reconstructing pile-up events acquired with a neutron detector array (NEDA) using an one-dimensional convolutional autoencoder (1D-CAE). The datasets for training and testing the 1D-CAE are created from data acquired from the NEDA. The new pile-up signal reconstruction method is evaluated from the point of view of how similar the reconstructed signals are to the original ones. Furthermore, it is analysed considering the result of the neutron-gamma discrimination based on charge comparison, comparing the result obtained from original and reconstructed signals.
|
|
|
Ferrer-Sanchez, A., Martin-Guerrero, J., Ruiz de Austri, R., Torres-Forne, A., & Font, J. A. (2024). Gradient-annihilated PINNs for solving Riemann problems: Application to relativistic hydrodynamics. Comput. Meth. Appl. Mech. Eng., 424, 116906–18pp.
Abstract: We present a novel methodology based on Physics-Informed Neural Networks (PINNs) for solving systems of partial differential equations admitting discontinuous solutions. Our method, called Gradient-Annihilated PINNs (GA-PINNs), introduces a modified loss function that forces the model to partially ignore high-gradients in the physical variables, achieved by introducing a suitable weighting function. The method relies on a set of hyperparameters that control how gradients are treated in the physical loss. The performance of our methodology is demonstrated by solving Riemann problems in special relativistic hydrodynamics, extending earlier studies with PINNs in the context of the classical Euler equations. The solutions obtained with the GA-PINN model correctly describe the propagation speeds of discontinuities and sharply capture the associated jumps. We use the relative l(2) error to compare our results with the exact solution of special relativistic Riemann problems, used as the reference ''ground truth'', and with the corresponding error obtained with a second-order, central, shock-capturing scheme. In all problems investigated, the accuracy reached by the GA-PINN model is comparable to that obtained with a shock-capturing scheme, achieving a performance superior to that of the baseline PINN algorithm in general. An additional benefit worth stressing is that our PINN-based approach sidesteps the costly recovery of the primitive variables from the state vector of conserved variables, a well-known drawback of grid-based solutions of the relativistic hydrodynamics equations. Due to its inherent generality and its ability to handle steep gradients, the GA-PINN methodology discussed in this paper could be a valuable tool to model relativistic flows in astrophysics and particle physics, characterized by the prevalence of discontinuous solutions.
|
|
|
Ferrer-Sanchez, A., Villanueva-Espinosa, N., Hernani-Morales, C., Ruiz de Austri-Bazan, R., Font, J. A., Martin-Guerrero, J. D., et al. (2026). Addressing the gravitational collapse of a massless scalar field with physics-informed neural networks. Mach. Learn.-Sci. Technol., 7(2), 025038–26pp.
Abstract: The gravitational collapse of a massless scalar field remains a demanding benchmark for numerical methods in numerical relativity, as it exhibits critical behavior at the boundary between dispersion and black hole formation. In this work, we revisit this problem by relying on physics-informed neural networks (PINNs) as flexible solvers for partial differential equations, thereby providing a comparative assessment of several recent neural architectures. Building on the Einstein-massless-Klein-Gordon formulation in polar-areal coordinates, we consider four initial-value problems encompassing subcritical, critical, and supercritical regimes and use high-resolution finite-difference simulations as reference solutions. Our study is primarily comparative: we evaluate several state-of-the-art deep learning architectures, including vanilla and high-precision PINNs, sinusoidal-feature and quadratic-residual variants, and Kolmogorov-Arnold networks, all trained under a common loss design that encodes the field equations, boundary conditions, and causal time-space enforcement, together with a novel adaptive spacetime sampling. Within this framework, we also introduce ModPINN, a modest modification of standard PINNs that augments standard multilayer perceptrons with coordinate embeddings, quadratic layers, and other common ingredients in recent literature. This study shows that deep-learning-based methods can reproduce finite-difference solutions for the scalar field and the spacetime metric with competitive accuracy using significantly fewer collocation points than more traditional methodologies. While no single architecture dominates in all regimes, ModPINN achieves particularly stable and accurate solutions near criticality, indicating that suitably designed embeddings and adaptive sampling can enhance the robustness of PINNs for challenging gravitational-collapse scenarios.
|
|
|
Folgado, M. G., & Sanz, V. (2022). Exploring the political pulse of a country using data science tools. J. Comput. Soc. Sci., 5, 987–1000.
Abstract: In this paper we illustrate the use of Data Science techniques to analyse complex human communication. In particular, we consider tweets from leaders of political parties as a dynamical proxy to political programmes and ideas. We also study the temporal evolution of their contents as a reaction to specific events. We analyse levels of positive and negative sentiment in the tweets using new tools adapted to social media. We also train a Fully-Connected Neural Network (FCNN) to recognise the political affiliation of a tweet. The FCNN is able to predict the origin of the tweet with a precision in the range of 71-75%, and the political leaning (left or right) with a precision of around 90%. This study is meant to be viewed as an example of how to use Twitter data and different types of Data Science tools for a political analysis.
|
|
|
Folgado, M. G., Sanz, V., Hirn, J., Lorenzo-Saez, E., & Urchueguia, J. F. (2025). Towards Predictive Pollution Control Through Traffic Flux Forecasting With Deep Learning: A Case Study in the City of Valencia. Applied AI Lett., 6(1), e106–15pp.
Abstract: Traffic congestion represents a significant urban challenge, with notable implications for public health and environmental well-being. Consequently, urban decision-makers prioritize the mitigation of congestion. This study delves into the efficacy of harnessing extensive data on urban traffic dynamics, coupled with comprehensive knowledge of road networks, to enable Artificial Intelligence (AI) in forecasting traffic flux well in advance. Such forecasts hold promise for informing emission reduction measures, particularly those aligned with Low Emission Zone policies. The investigation centers on Valencia, leveraging its robust traffic sensor infrastructure, one of the most densely deployed worldwide, encompassing approximately 3500 sensors strategically positioned across the city. Employing historical data spanning 2016 and 2017, we undertake the task of training and characterizing a Long Short-Term Memory (LSTM) Neural Network for the prediction of temporal traffic patterns. Our findings demonstrate the LSTM's efficacy in real-time forecasting of traffic flow evolution, facilitated by its ability to discern salient patterns within the dataset.
|
|
|
Khosa, C. K., Mars, L., Richards, J., & Sanz, V. (2020). Convolutional neural networks for direct detection of dark matter. J. Phys. G, 47(9), 095201–20pp.
Abstract: The XENON1T experiment uses a time projection chamber (TPC) with liquid xenon to search for weakly interacting massive particles (WIMPs), a proposed dark matter particle, via direct detection. As this experiment relies on capturing rare events, the focus is on achieving a high recall of WIMP events. Hence the ability to distinguish between WIMP and the background is extremely important. To accomplish this, we suggest using convolutional neural networks (CNNs); a machine learning procedure mainly used in image recognition tasks. To explore this technique we use XENON collaboration open-source software to simulate the TPC graphical output of dark matter signals and main backgrounds. A CNN turns out to be a suitable tool for this purpose, as it can identify features in the images that differentiate the two types of events without the need to manipulate or remove data in order to focus on a particular region of the detector. We find that the CNN can distinguish between the dominant background events (ER) and 500 GeV WIMP events with a recall of 93.4%, precision of 81.2% and an accuracy of 87.2%.
|
|
|
Perez-Curbelo, J., Roser, J., Muñoz, E., Barrientos, L., Sanz, V., & Llosa, G. (2025). Improving Compton camera imaging of multi-energy radioactive sources by using machine learning algorithms for event selection. Radiat. Phys. Chem., 226, 112166–11pp.
Abstract: Event selection and background reduction for Compton camera imaging of multi-energy radioactive sources has been performed by employing neural networks. A Compton camera prototype with detectors made of LaBr3 crystals coupled to silicon photomultiplier arrays was used to acquire experimental data from a circular array of Na-22 sources. The prototype and two arrays of Na-22 sources were simulated with GATE v8.2 Monte Carlo code, to obtain data for neural network training. Neural network models were trained on simulated data for event classification. The optimum models were found by using Weights & Biases platform tools. The trained models were used to classify simulated and real data for selecting signal events and rejecting background prior to image reconstruction. The models performed well on simulated data. The image obtained with experimental data showed an improvement with respect to event selection with energy cuts. The method is promising for Compton camera imaging of multi-energy radioactive sources.
|
