arXiv subjects

Nicola Serra

Publications and source records attributed to Nicola Serra.

At least 19 recordsLinked to original sources

arXiv2026

RAISE: LLM-based Automated Heuristic Design with Robust Adversary Instance Search

Automated Heuristic Design (AHD) with Large Language Models (LLMs) has shown remarkable progress in discovering high-quality heuristics. However, existing LLM-based AHD methods optimize heuristics for a fixed training instance set and may fail catastrophically when deployed under real-world distributional shifts. We propose Robust Adversary Instance Search (RAISE), a framework that integrates constrained worst-case instance search within a principled neighborhood of the training distribution into the LLM-based evolutionary search loop. RAISE treats robust AHD as a constrained adversarial instance search problem: the outer loop evolves heuristics via LLM operators, while an LLM-free inner loop efficiently identifies hard instances within an epsilon-ball around the training instance set using a basis distribution parameterization with boundary projection. Comprehensive experiments on Online Bin Packing (OBP), Online Job Shop Scheduling (OJSP), and Online Vehicle Routing (OVRP) across five distribution families demonstrate that existing LLM-based AHD methods degrade by up to 19 times under distribution shift, while RAISE consistently maintains strong performance across all tested distributions and problem scales

cs.AI↗

arXiv2026

Joint Treatment Effect Estimation from Incomplete Healthcare Data: Temporal Causal Normalizing Flows with LLM-driven Evolutionary MNAR Imputation

Target trial emulation (TTE) enables causal questions to be studied with observational data when randomized controlled trials (RCTs) are infeasible. Yet treatment-effect methods often address causal estimation, missingness, and temporal structure separately, limiting their robustness in electronic health records (EHRs), where time-varying confounding and missing-not-at-random (MNAR) biomarkers can reach 50%--80%. We propose a two-stage pipeline for treatment effect estimation from incomplete longitudinal EHRs. First, CausalFlow-T, a directed acyclic graph (DAG)-constrained normalizing flow with long short-term memory (LSTM)-encoded patient history, performs exact invertible counterfactual inference, avoiding approximation errors from variational inference and separating confounding through explicit causal structure. Ablations on four synthetic and one semi-synthetic benchmark with known counterfactuals show that DAG constraints and exact inference address distinct failure modes: neither compensates for the other. Second, because CausalFlow-T requires completed inputs, we introduce an LLM-driven evolutionary imputer that proposes executable imputation operators rather than individual entries, and evaluate it with three large language model (LLM) backends, including two open-source models. Across 30%--80% MNAR missingness, this imputer achieves the best pooled rank over biomarker and causal metrics, leading in point-wise accuracy and temporal extrapolation while preserving average treatment effect (ATE) recovery as statistical baselines degrade. On Swiss primary-care EHRs from adults with type 2 diabetes initiating a GLP-1 receptor agonist or SGLT-2 inhibitor, the pipeline estimates a per-protocol weight-loss difference of -0.98 kg [95% CI -1.01, -0.96] favoring GLP-1 receptor agonists, consistent with randomized evidence and obtained from realistically incomplete real-world EHRs.

cs.LG↗

arXiv2026

On the Codesign of Scientific Experiments and Industrial Systems

The optimization of large experiments in fundamental science, such as detectors for subnuclear physics at particle colliders, shares with the optimization of complex systems for industrial or societal applications the common issue of addressing the inter-relation between parameters describing the hardware used in data production and parameters used to analyse those data. While in many cases this coupling can be ignored -- when the problem can be successfully factored into simpler sub-tasks and the latter addressed serially -- there are situations in which that approach fails to converge to the absolute maximum of expected performance, as it results in a mis-alignment of the optimized hardware and software solutions. In this work we consider a few use cases of interest in fundamental science collected primarily from particle physics and related areas, and a pot-pourri of industrial and societal applications where the matter is similarly of relevance. We discuss the emergence of strong hardware-software coupling in some of those systems, as well as co-design procedures that may be deployed to identify the global maximum of their relevant utility functions. We observe how numerous opportunities exist to advance methods and tools for hardware-software co-design optimization, bridging fundamental science and industry through application- and challenge-driven projects, and shaping the future of scientific experiments and industrial systems.

physics.ins-det↗

arXiv2026

Towards replacing detector simulation with heterogeneous GNNs in flavour physics analyses

Driven by the increasing volume of recorded data, the demand for simulation from experiments based at the Large Hadron Collider will rise sharply in the coming years. Addressing this demand solely with existing computationally intensive workflows is not feasible. This paper introduces a new fast simulation tool designed to address this demand at the LHCb experiment. This tool emulates the detector response to arbitrary multibody decay topologies at LHCb. Rather than memorising specific decay channels, the model learns generalisable patterns within the response, allowing it to interpolate to channels not present in the training data. Novel heterogeneous graph neural network architectures are employed that are designed to embed the physical characteristics of the task directly into the network structure. We demonstrate the performance of the tool across a range of decay topologies, showing the networks can correctly model the relationships between complex variables. The architectures and methods presented are generic and could readily be adapted to emulate workflows at other simulation-intensive particle physics experiments.

hep-ex↗

arXiv2026

Large Language Models for Physics Instrument Design

We study the use of large language models (LLMs) for physics instrument design and compare their performance to reinforcement learning (RL). Using only prompting, LLMs are given task constraints and summaries of prior high-scoring designs and propose complete detector configurations, which we evaluate with the same simulators and reward functions used in RL-based optimization. Although RL yields stronger final designs, we find that modern LLMs consistently generate valid, resource-aware, and physically meaningful configurations that draw on broad pretrained knowledge of detector design principles and particle--matter interactions, despite having no task-specific training. Based on this result, as a first step toward hybrid design workflows, we explore pairing the LLMs with a dedicated trust region optimizer, serving as a precursor to future pipelines in which LLMs propose and structure design hypotheses while RL performs reward-driven optimization. Based on these experiments, we argue that LLMs are well suited as meta-planners: they can design and orchestrate RL-based optimization studies, define search strategies, and coordinate multiple interacting components within a unified workflow. In doing so, they point toward automated, closed-loop instrument design in which much of the human effort required to structure and supervise optimization can be reduced.

physics.ins-det↗

arXiv2025

Ultra-Fast Muon Transport via Histogram Sampling on GPUs

We present a GPU-accelerated method for muon transport based on histogram sampling that delivers orders of magnitude faster performance than CPU-based Geant4 simulation. Our method employs precomputed histograms of momentum loss and scattering, derived from detailed Geant4 simulations, to statistically reproduce all the non-decaying physics processes during muon traversal through matter. Implemented as a CUDA kernel, the parallel algorithm enables the concurrent simulation of tens of thousands of particles on a single GPU whilst taking into account a complex geometry and a magnetic field force integrated using a fourth-order Runge-Kutta method. Validation against Geant4 in both simple and realistic detector geometries shows that the approach preserves key physical features while achieving speedups of several orders of magnitude, even compared to CPU-based simulations on a large CPU farm with over a thousand cores. This work highlights the significant potential of GPU-based implementations for particle transport, with applicability extending to neutrino propagation and future implementations including discrete processes such as particle decay.

physics.comp-ph↗

arXiv2025

Leveraging Reinforcement Learning, Genetic Algorithms and Transformers for background determination in particle physics

Experimental studies of beauty hadron decays face significant challenges due to a wide range of backgrounds arising from the numerous possible decay channels with similar final states. For a particular signal decay, the process for ascertaining the most relevant background processes necessitates a detailed analysis of final state particles, potential misidentifications, and kinematic overlaps, which, due to computational limitations, is restricted to the simulation of only the most relevant backgrounds. Moreover, this process typically relies on the physicist's intuition and expertise, as no systematic method exists. This paper has two primary goals. First, from a particle physics perspective, we present a novel approach that utilises Reinforcement Learning (RL) to overcome the aforementioned challenges by systematically determining the critical backgrounds affecting beauty hadron decay measurements. While beauty hadron physics serves as the case study in this work, the proposed strategy is broadly adaptable to other types of particle physics measurements. Second, from a Machine Learning perspective, we introduce a novel algorithm which exploits the synergy between RL and Genetic Algorithms (GAs) for environments with highly sparse rewards and a large trajectory space. This strategy leverages GAs to efficiently explore the trajectory space and identify successful trajectories, which are used to guide the RL agent's training. Our method also incorporates a transformer architecture for the RL agent to handle token sequences representing decays.

cs.LG↗

arXiv2025

Scalable Multi-Task Learning for Particle Collision Event Reconstruction with Heterogeneous Graph Neural Networks

The growing luminosity frontier at the Large Hadron Collider is challenging the reconstruction and analysis of particle collision events. Increased particle multiplicities are straining latency and storage requirements at the data acquisition stage, while new complications are emerging, including higher background levels and more frequent particle vertex misassociations. This in turn necessitates the development of more holistic and scalable reconstruction methods that take advantage of recent advances in machine learning. We propose a novel Heterogeneous Graph Neural Network (HGNN) architecture featuring unique representations for diverse particle collision relationships and integrated graph pruning layers for scalability. Trained with a multi-task paradigm in an environment mimicking the LHCb experiment, this HGNN significantly improves beauty hadron reconstruction performance. Notably, it concurrently performs particle vertex association and graph pruning within a single framework. We quantify reconstruction and pruning performance, demonstrate enhanced inference time scaling with event complexity, and mitigate potential performance loss using a weighted message passing scheme.

physics.data-an↗

arXiv2024

Physics Instrument Design with Reinforcement Learning

We present a case for the use of Reinforcement Learning (RL) for the design of physics instrument as an alternative to gradient-based instrument-optimization methods. It's applicability is demonstrated using two empirical studies. One is longitudinal segmentation of calorimeters and the second is both transverse segmentation as well longitudinal placement of trackers in a spectrometer. Based on these experiments, we propose an alternative approach that offers unique advantages over differentiable programming and surrogate-based differentiable design optimization methods. First, Reinforcement Learning (RL) algorithms possess inherent exploratory capabilities, which help mitigate the risk of convergence to local optima. Second, this approach eliminates the necessity of constraining the design to a predefined detector model with fixed parameters. Instead, it allows for the flexible placement of a variable number of detector components and facilitates discrete decision-making. We then discuss the road map of how this idea can be extended into designing very complex instruments. The presented study sets the stage for a novel framework in physics instrument design, offering a scalable and efficient framework that can be pivotal for future projects such as the Future Circular Collider (FCC), where most optimized detectors are essential for exploring physics at unprecedented energy scales.

physics.ins-det↗

arXiv2024

Human-in-the-loop Reinforcement Learning for Data Quality Monitoring in Particle Physics Experiments

Data Quality Monitoring (DQM) is a crucial task in large particle physics experiments, since detector malfunctioning can compromise the data. DQM is currently performed by human shifters, which is costly and results in limited accuracy. In this work, we provide a proof-of-concept for applying human-in-the-loop Reinforcement Learning (RL) to automate the DQM process while adapting to operating conditions that change over time. We implement a prototype based on the Proximal Policy Optimization (PPO) algorithm and validate it on a simplified synthetic dataset. We demonstrate how a multi-agent system can be trained for continuous automated monitoring during data collection, with human intervention actively requested only when relevant. We show that random, unbiased noise in human classification can be reduced, leading to an improved accuracy over the baseline. Additionally, we propose data augmentation techniques to deal with scarce data and to accelerate the learning process. Finally, we discuss further steps needed to implement the approach in the real world, including protocols for periodic control of the algorithm's outputs.

hep-ex↗

arXiv2024

FunTuple: A new N-tuple component for offline data processing at the LHCb experiment

The offline software framework of the LHCb experiment has undergone a significant overhaul to tackle the data processing challenges that will arise in the upcoming Run 3 and Run 4 of the Large Hadron Collider. This paper introduces FunTuple, a novel component developed for offline data processing within the LHCb experiment. This component enables the computation and storage of a diverse range of observables for both reconstructed and simulated events by leveraging on the tools initially developed for the trigger system. This feature is crucial for ensuring consistency between trigger-computed and offline-analysed observables. The component and its tool suite offer users flexibility to customise stored observables, and its reliability is validated through a full-coverage set of rigorous unit tests. This paper comprehensively explores FunTuple's design, interface, interaction with other algorithms, and its role in facilitating offline data processing for the LHCb experiment for the next decade and beyond.

physics.data-an↗

arXiv2023

The DL Advocate: Playing the devil's advocate with hidden systematic uncertainties

We propose a new method based on machine learning to \emph{play the devil's advocate} and investigate the impact of unknown systematic effects in a quantitative way. This method proceeds by reversing the measurement process and using the physics results to interpret systematic effects under the Standard Model hypothesis. We explore this idea with two alternative approaches: the first one relies on a combination of gradient descent and optimisation techniques, its application and potentiality is illustrated with an example that studies the branching fraction measurement of a heavy-flavour decay. The second method employs reinforcement learning and it is applied to the determination of the $P_{5}^{'}$ angular observable in $B^0 \to K^{*0} μ^+μ^-$ decays. We find that for the former, the size of a hypothetical hidden systematic uncertainty strongly depends on the kinematic overlap between the signal and normalisation channel, while the latter is very robust against possible mismodellings of the efficiency.

hep-ex↗

arXiv2023

GNN for Deep Full Event Interpretation and hierarchical reconstruction of heavy-hadron decays in proton-proton collisions

The LHCb experiment at the Large Hadron Collider (LHC) is designed to perform high-precision measurements of heavy-hadron decays, which requires the collection of large data samples and a good understanding and suppression of multiple background sources. Both factors are challenged by a five-fold increase in the average number of proton-proton collisions per bunch crossing, corresponding to a change in the detector operation conditions for the LHCb Upgrade I phase, recently started. A further ten-fold increase is expected in the Upgrade II phase, planed for the next decade. The limits in the storage capacity of the trigger will bring an inverse relation between the amount of particles selected to be stored per event and the number of events that can be recorded, and the background levels will raise due to the enlarged combinatorics. To tackle both challenges, we propose a novel approach, never attempted before in a hadronic collider: a Deep-learning based Full Event Interpretation (DFEI), to perform the simultaneous identification, isolation and hierarchical reconstruction of all the heavy-hadron decay chains per event. This approach radically contrasts with the standard selection procedure used in LHCb to identify heavy-hadron decays, that looks individually at sub-sets of particles compatible with being products of specific decay types, disregarding the contextual information from the rest of the event. We present the first prototype for the DFEI algorithm, that leverages the power of Graph Neural Networks (GNN). This paper describes the design and development of the algorithm, and its performance in Upgrade I simulated conditions.

hep-ex↗

arXiv2023

Universal approximation of credit portfolio losses using Restricted Boltzmann Machines

We introduce a new portfolio credit risk model based on Restricted Boltzmann Machines (RBMs), which are stochastic neural networks capable of universal approximation of loss distributions. We test the model on an empirical dataset of default probabilities of 1'012 US companies and we show that it outperforms commonly used parametric factor copula models -- such as the Gaussian or the t factor copula models -- across several credit risk management tasks. In particular, the model leads to better fits for the empirical loss distribution and more accurate risk measure estimations. We introduce an importance sampling procedure which allows risk measures to be estimated at high confidence levels in a computationally efficient way and which is a substantial improvement over the Monte Carlo techniques currently available for copula models. Furthermore, the statistical factors extracted by the model admit an interpretation in terms of the underlying portfolio sector structure and provide practitioners with quantitative tools for the management of concentration risk. Finally, we show how to use the model for stress testing by estimating stressed risk measures (e.g. stressed VaR) under various macroeconomic stress test scenarios, such as those specified by the FRB's Dodd-Frank Act stress test.

q-fin.CP↗

arXiv2021

A general effective field theory description of $b \to s l^+ l^-$ lepton universality ratios

We construct an expression for a general lepton flavour universality (LFU) ratio, $R_{X}$, in $b\to s l^+ l^-$ decays in terms of a series of hadronic quantities which can be treated as nuisance parameters. This expression allows to include any LFU ratio in global fits of $b\to s l^+ l^-$ short-distance parameters, even in the absence of a precise knowledge of the corresponding hadronic structure. The absence of sizeable LFU violation and the approximate left-handed structure of the Standard Model amplitude imply that only a very limited set of hadronic parameters hamper the sensitivity of $R_X$ to a possible LFU violation of short-distance origin. A global $b\to s l^+ l^-$ combination is performed including the measurement of $R_{pK}$ for the first time, resulting in a significance of new physics of $4.2\,σ$. In light of this, we evaluate the impact on the global significance of new physics using a set of experimentally promising non-exclusive $R_X$ measurements that LHCb can perform, and find that they can significantly increase the discovery potential of the experiment.

hep-ph↗

arXiv2021

On the significance of new physics in $b\to s\ell^+\ell^-$ decays

Motivated by deviations with respect to Standard Model predictions in $b\to s\ell^+\ell^-$ decays, we evaluate the global significance of the new physics hypothesis in this system by including the {\it look-elsewhere effect} for the first time. We estimate the trial-factor with pseudo-experiments and find that it can be as large as seven. We calculate the global significance for the new physics hypothesis by considering the most general description of a non-standard $b\to s\ell^+\ell^-$ amplitude of short-distance origin. Theoretical uncertainties are treated in a highly conservative way by absorbing the corresponding effects into a redefinition of the Standard Model amplitude. Using the most recent measurements of LHCb, ATLAS and CMS, we obtain the global significance to be $4.3$ standard deviations.

hep-ph↗

arXiv2021

A High Pressure Time Projection Chamber with Optical Readout

Measurements of proton-nucleus scattering and high resolution neutrino-nucleus interaction imaging are key to reduce neutrino oscillation systematic uncertainties in future experiments. A High Pressure Time Projection Chamber (HPTPC) prototype has been constructed and operated at Royal Holloway University of London and CERN as a first step in the development of a HPTPC capable of performing these measurements as part of a future long-baseline neutrino oscillation experiment such as the Deep Underground Neutrino Experiment. In this paper we describe the design and operation of the prototype HPTPC with an argon based gas mixture. We report on the successful hybrid charge and optical readout, using four CCD cameras, of signals from Am-241 sources.

physics.ins-det↗

arXiv2020

zfit: scalable pythonic fitting

Statistical modeling is a key element in many scientific fields and especially in High-Energy Physics (HEP) analysis. The standard framework to perform this task in HEP is the C++ ROOT/RooFit toolkit; with Python bindings that are only loosely integrated into the scientific Python ecosystem. In this paper, zfit, a new alternative to RooFit written in pure Python, is presented. Most of all, zfit provides a well defined high-level API and workflow for advanced model building and fitting, together with an implementation on top of TensorFlow, allowing a transparent usage of CPUs and GPUs. It is designed to be extendable in a very simple fashion, allowing the usage of cutting-edge developments from the scientific Python ecosystem in a transparent way. The main features of zfit are introduced, and its extension to data analysis, especially in the context of HEP experiments, is discussed.

physics.data-an↗