Publications

2026

• Convergence of stochastic particle systems and applications
  
  • Priser Victor
  , 2026. This thesis investigates the theoretical behavior and practical applications of stochastic particle systems, with a focus on their convergence properties and their use in optimization, simulation, Monte Carlo methods, and variational inference.The thesis is organized around two main classes of systems: McKean-Vlasov particle systems and Importance Sampling particle systems.The first part studies McKean-Vlasov systems, where particles interact through their collective distribution. Rather than considering the idealized infinite-particle limit, the thesis focuses on the more realistic setting of a finite number of particles and analyzes their long-time behavior. It shows that the empirical distribution converges to the set of stationary solutions of the associated McKean-Vlasov equation under mild assumptions. These results are then applied to two key algorithms: Consensus-Based Optimization (CBO), for which convergence toward the global minimizer is established with explicit rates, and Stein Variational Gradient Descent (SVGD), for which a noisy variant is introduced to guarantee convergence to the target distribution.The second part addresses Importance Sampling systems, where weighted particles approximate a target distribution. To tackle the issue of weight degeneracy, a kernel-based adaptive method is proposed to reduce weight variance while preserving convergence. The thesis also introduces a new gradient-free optimization approach based on weighted averages, which outperforms standard random search methods while maintaining similar computational complexity.Overall, the thesis provides new theoretical insights into the long-time behavior of stochastic particle systems and proposes improved algorithms for sampling and optimization, supported by rigorous convergence guarantees.
• SPOT : un jeu de données français annoté pour la détection d’interventions critiques dans les conversations en ligne
  
  • Berriche Manon
  • Nouri Célia
  • Clavel Chloé
  • Cointet Jean-Philippe
  , 2026, pp.45-46. Nous présentons SPOT (Stopping Points in Online Threads), le premier corpus annoté traduisant le concept de sociologie pragmatique de « point d'arrêt » en une tâche reproductible de traitement automatique du langage naturel (TALN). Les points d'arrêt sont des interventions critiques ordinaires qui interrompent ou redirigent les discussions en ligne sous diverses formes — ironie, doute subtil ou arguments fragmentaires. Nous rendons opérationnel ce concept sous la forme d'une tâche de classification binaire, et fournissons un guide d'annotation précis pour cette tâche menant à des scores d'accords inter-annotateurs robustes. Le corpus contient 43 305 commentaires Facebook en français annotés manuellement comme “point d’arrêt” ou non, et enrichis avec des métadonnées contextuelles (article partagé, publication, commentaire parent, page ou groupe, et source médiatique). Ces commentaires sont liés à des publications partageant des URL signalées comme étant des fausses informations par les utilisateurs sur des groupes ou pages Facebook publiques. Nous comparons des modèles d'encodeurs affinés (CamemBERT) et des grands modèles de langue (LLM) promptés avec différentes stratégies de prompting. Les résultats montrent que les encodeurs affinés surpassent les LLM promptés de plus de 10 points de pourcentage en termes de score F1, ce qui confirme l'importance de l'apprentissage supervisé de tâches sociales et interactionnelles nouvelles ou peu documentées, comme la détection des points d’arrêt dans des commentaires Facebook, notamment lorsque ces tâches sont non anglophones. L'intégration de métadonnées contextuelles améliore encore les scores F1 des modèles d'encodeurs, qui passent de 0,75 à 0,78. Nous publions l'ensemble de données anonymisées, ainsi que le guide d'annotation et notre code (https://anonymous.4open.science/r/SPOT-benchmark-2E65), afin de favoriser la transparence et la reproductibilité de la recherche.
• MEDIATE: multi-faceted implementation of a mixed software/hardware-based zero trust framework for the computing continuum
  
  • Fournaris Apostolos P
  • Haleplidis Evangelos
  • Abdoul-Soukour Shahin
  • Huang Chih-Kai
  • Khoder Niemat
  • Bouloukakis Georgios
  • Brokalakis Andreas
  • Georgopoulos Konstantinos
  • Ioannidis Sotiris
  , 2026. This paper introduces MEDIATE, a framework that brings zero trust principles to the Cloud-Edge-IoT Computing Continuum. MEDIATE integrates several software-based security and policy services operating at the cloud and edge layers with hardware-and software-based monitoring components as IoT devices deployed close to the assets that require security protection. The framework is designed to a three-layer architecture with Orchestrator (Cloud), Overwatch (Edge), and Sentinel (IoT) that collectively enables cross-layer policy enforcement and coordinated threat mitigation. The Orchestrator provides global view and policy consistency, the Overwatch performs local analysis and threat correlation, and the Sentinel acts as lightweight agent for real-time monitoring near critical assets. These components can provide an adaptive security foundation suitable for distributed and large-scale operational domains such as Fourth-Party Logistics (4PL).
• Identifying and Evaluating Misleading Climate Communication with Natural Language Processing
  
  • Calamai Tom
  , 2026. Climate communication is becoming more abundant, but not necessarily more informative. This thesis investigates whether Natural Language Processing (NLP) can help structure climate-related discourse and distinguish substantive content from vague or rhetorical formulations. By examining the literature on greenwashing and major datasets for climate-related NLP tasks, it highlights key limitations, including subjectivity, ambiguity, and noisy data. It then proposes ways to address these issues through annotation schemes and evaluation metrics designed for ambiguity, as well as methods for propagating uncertainty into downstream analyses. Overall, the thesis shows that NLP can make climate-related discourse more explicit and analyzable.
• RV-Sec5: Enhancing Pre-Silicon Security Evaluation of RISC-V Processors through Targeted ISA-Level Instrumentation in the gem5 Simulation Framework
  
  • Awais Muhammad
  • Mushtaq Maria
  • Naviner Lirida
  • Yahya Jawad Haj
  • Bruguier Florent
  Pragmatic Cybersecurity, sciltp, 2026, 1 (1), pp.1-25. The open-source RISC-V Instruction Set Architecture (ISA) is being adopted rapidly in security-sensitive areas such as IoT, edge computing, and aerospace systems, which makes early-stage security validation increasingly important. Yet most existing approaches still depend either on post-silicon testing or on high-level emulation. Neither is well suited to exposing ISA-specific vulnerabilities or microarchitectural side effects during the design phase. As a result, there remains a gap between highlevel security policies and the way hardware actually behaves at runtime, and that gap can leave processors exposed to privilege escalation, memory protection failures, and side-channel leakage that may only become visible late in development. In this paper, we present RV-Sec5, a systematic and policy-driven framework for ISA-level security evaluation built on the gem5 cycle-accurate simulator. RV-Sec5 provides a formal method for translating high-level security invariants, including privilege isolation, Physical Memory Protection (PMP) enforcement, and Control and Status Register (CSR) integrity, into automated cycle-accurate instrumentation points embedded directly in the ISA decoder. By recording precise architectural execution context at instruction commit time, the framework supports specification-driven methodology of privilege escalation attempts and enables systematic correlation between ISA-level events and microarchitectural behavior, including TLB activity and cache state changes, without interfering with functional execution. Our results shows that RV-Sec5 can significantly detects the specification violation on the events that are permitted in User mode, the results shows that after extending the gem5 with the hooks added and ISA modified, it create an overhead on the simulation. The overall overhead of RV-Sec5 for the simulation time is less than 4% and the overhead for the memory usage is less than 2% across the evaluated workloads. RV-Sec5 is a modular, cycle-accurate observation and post-execution detection virtual platform that reduces the gap between architectural security requirements and their enforcement at the Microarchitectural level, using post-silicon testing within the RISC-V processor design flow. (10.53941/pc.2026.100006)
  DOI : 10.53941/pc.2026.100006
• Le corpus NeWMe : un corpus de référence pour l’étude de la négociation du sens des mots
  
  • Soler Aina Garí
  • Myrendal Jenny
  • Clavel Chloé
  • Larsson Staffan
  , 2026, pp.27-28. Les séquences de négociation du sens des mots (*Word Meaning Negotiation*, WMN) apparaissent lorsque des interlocuteurs se concentrent sur la clarification ou la négociation du sens d’un mot ou d’une expression. Ces interactions orientent temporairement la conversation vers l'exploration des nuances de sens, menant à des clarification rapides ou à des débats plus approfondis. Cet article présente le plus grand corpus de WMN librement accessible annoté manuellement. Il comprend des conversations orales dyadiques et multipartites ainsi que des débats en ligne. Notre méthodologie combine une recherche par expressions régulières avec un schéma d’annotation distinguant plusieurs types de WMN, notamment celles déclenchées par une non-compréhension (NONs) ou un désaccord (DINs), ainsi que des phénomènes connexes. Les analyses révèlent des tendances selon les contextes : les NONs sont plus fréquentes à l’oral, tandis que les DINs dominent en ligne. Cette ressource contribue à l’étude de l’alignement sémantique et à l’identification des WMN.
• Analyse Automatique de la Collaboration à partir de Ressources de Données Conversationnelles Humaines : Une Revue
  
  • Yu Yi
  • Boritchev Maria
  • Clavel Chloé
  , 2026, pp.2-2. La collaboration entre êtres humains, dans le but d'accomplir une tâche commune, met en jeu des interactions comportementales complexes. Une partie de ces interactions s'explicite à travers les conversations, qui constitue le principal moyen d’échanger des informations et de se coordonner. Les données conversationnelles constituent ainsi une ressource précieuse pour l’analyse automatique des processus collaboratifs. Dans cet article, nous nous concentrons sur les interactions verbales et proposons une analyse de la collaboration à partir de ressources conversationnelles portant sur la réalisation de tâches précises. Nous couvrons dans cet article les théories linguistiques associées, les schémas d'annotation, les différents types de tâches considérées ainsi que différentes approches de modélisation. Notre objectif est d’examiner comment exploiter des données conversationnelles humaines pour l’analyse de la collaboration. Nous espérons que cette analyse constituera une ressource pratique et qu’elle permet de mettre en lumière des pistes encore peu explorées, en vue de travaux futurs.
• Affinés pour la réussite : Évaluation des représentations dans le domaine de la formation professionnelle
  
  • Breidenstein Alicia
  • Leang Marguerite
  • Labeau Matthieu
  , 2026, pp.65-90. Dans le domaine de la formation professionnelle et du marché du travail, les données sont souvent structurées sous forme de graphes avec attributs textuels, reliant les offres de formation, les CV et les ontologies de certifications et de compétences. Cette configuration présente des défis uniques, car ces données sont organisées selon une hiérarchie de milliers d’étiquettes de classification, et le graphe comporte de nombreuses liaisons manquantes. Avec la difficulté supplémentaire que représente le vocabulaire spécifique au domaine, il est nécessaire d’utiliser une représentation du texte à la fois adaptée et efficace. Dans ce travail, nous évaluons un large éventail de représentations de texte, allant des représentations symboliques aux grands modèles de langue, sur des tâches réelles appliquées à notre jeu de données interne. Nous montrons que les représentations lexicales offrent les meilleures performances sans affinage, mais que les modèles basés sur BERT dominent une fois affinés sur des données spécifiques au domaine. En revanche, les LLM utilisés avec des instructions génératives directes sous-performent, limités par la complexité structurelle des données et le vocabulaire spécifique, et atteignent des performances comparables à celles des modèles basés sur BERT lorsqu'ils sont affinés sur le jeu de données.
• Efficient 3D Deep Learning for Joint DBT Reconstruction and Faithful Uncertainty Estimation: Preliminary Clinical Evaluation
  
  • Quillent Arnaud
  • Bismuth Vincent
  • Bloch Isabelle
  • Kervazo Christophe
  • Ladjal Saïd
  , 2026. Purpose: Digital breast tomosynthesis (DBT) reconstruction is an ill-posed inverse problem due to the limited-angle acquisition and sparse projection sampling, leading to streak artefacts and poor depth resolution in the reconstructed volumes. Although deep learning provides data-driven priors, its application remains hindered by the lack of ground-truth clinical data and concerns regarding model reliability. This study introduces a 3D deep learning framework trained on virtual phantoms to jointly reconstruct DBT volumes and estimate voxel-wise uncertainty. Methods: We generate a synthetic database of phantom-reconstruction pairs derived from breast computed tomography (CT) images on which we train a post-processing 3D deep ensemble to estimate the reconstructed volume. To ensure reliability, we implement faithful regression and decouple the estimation of uncertainty from the main reconstruction task, mitigating training instability. The approach is then evaluated on the synthetic test set along with textured phantoms and clinical images. Results: On synthetic data, our method improves peak signal-to-noise ratio (PSNR) by 30%, normalised root mean squared error (NRMSE) by 50%, and structural similarity index measure (SSIM) by 43% compared to baseline iterative reconstruction. The qualitative analysis demonstrates significantly reduced tissue superposition and improved depth resolution in coronal and sagittal planes. Additionally, the uncertainty maps correlate well with the reconstruction error, but they slightly underestimate its true magnitude. A preliminary clinical evaluation yields plausible glandular localisation, although domain shift introduces some artefacts and texture mismatch. Lastly, we discuss the potential clinical value of uncertainty estimation as a tool to guide radiologists' confidence in reconstructed volumes. Conclusion: We successfully demonstrate a robust 3D deep learning approach for DBT that improves reconstruction quality while providing faithful uncertainty estimates. The model generalises well to synthetic phantoms, with promising translation to clinical images.
• Scenario-Based Learning Through and with AI: Evidence-Informed Simulations for Education Leaders
  
  • Hamonic Ella
  • Lugaz Candy
  • Demirag Annina
  • Sharrock Rémi
  • Thailinger Agustina
  • Picchio Maria Victoria
  • Swamy Vinitra
  • Mejia Paola
  , 2026, 3033, pp.145-152. (10.1007/978-3-032-29794-5_24)
  DOI : 10.1007/978-3-032-29794-5_24
• Towards Reliable and Secure RISC-V Systems: Survey of Testability and Security Mechanisms
  
  • Khan Mahreen
  • Mushtaq Maria
  • Apvrille Ludovic
  , 2026. <div><p>RISC-V has emerged as a versatile open-source instruction set architecture, enabling extensible microarchitectures, custom accelerators, and domain-specific processors. Its openness facilitates innovation in testability, safety, and security for safety-critical and security-sensitive applications. This survey provides a comprehensive review of recent research in RISC-V verification and protection mechanisms. We analyze AI-assisted test generation, statistical fault injection frameworks, systemlevel testing, design-for-test architectures, and hardware-software co-verification methods. In the safety domain, we discuss temporal isolation, performance monitoring, debug support, and fault containment strategies. Security mechanisms, including trusted execution environments, memory protection, cryptographic ISA extensions, post-quantum acceleration, and secure debug practices, are evaluated. Open challenges in scalable test coverage, AI-enabled certification, side-channel resilience, and lifecycle management are highlighted. Finally, we outline future research directions that leverage RISC-V's modularity and openness to enable trustworthy computing systems in automotive, aerospace, telecommunications, and edge computing domains.</p></div>
• Integrated photonics for coherent free-space optical communications established in the mid-infrared wavelength range
  
  • Nawfal Nour
  , 2026. This thesis investigates high-speed long-wavelength infrared (LWIR) photodetectors based on quantum cascade detector (QCD) architectures for free-space optical communication applications. While quantum cascade lasers have considerably expanded the possibilities for generating coherent radiation in the mid- and long-wave infrared, the realization of fast and sensitive receivers in this spectral range remains a major technological challenge.The work focuses on ridge-geometry QCDs operating around 9.6 µm. This geometry is also compatible, in the longer term, with integrated photonic approaches. The objective is to better understand the interplay between the intrinsic carrier dynamics of the quantum cascade active region and the extrinsic limitations introduced by device geometry and associated parasitic elements. A first generation of devices is investigated through a comprehensive characterization combining current-voltage, capacitance-voltage, responsivity, noise, electrical rectification, optical frequency response, and data-transmission demonstrations.The results show that the QCD active region is intrinsically compatible with high-speed operation, but that the measured bandwidth is initially strongly limited by parasitic capacitances associated with the coplanar access. To overcome this limitation, a second generation of devices was designed on semi-insulating InP:Fe substrates, with redesigned coplanar waveguides and dedicated de-embedding structures. This new architecture strongly reduces radio-frequency parasitics and provides experimental access to multi-gigahertz bandwidths governed by the detector itself.The thesis combines analytical modeling, full-wave electromagnetic simulations, and calibrated microwave measurements to separate the intrinsic behavior of the device from effects related to its electrical environment.
• D-Band Plastic Waveguide Design and Unified Channel Modeling for System-Level Simulation
  
  • Royer Louis
  • Mohellebi Reda
  • Vogt Lionel
  • Jabbour Chadi
  , 2026, pp.1-5. (10.1109/NewCAS64543.2026.11674008)
  DOI : 10.1109/NewCAS64543.2026.11674008
• Triboelectric-RF Hybrid Energy Harvesters: Lateral-Sliding and Contact-Separation Modes
  
  • Lu Yayi
  • Niotaki Kyriaki
  • Mohellebi Reda
  • Kalaia Karim Ben
  • Jabbour Chadi
  , 2026, pp.1-5. (10.1109/NewCAS64543.2026.11674045)
  DOI : 10.1109/NewCAS64543.2026.11674045
• Frequency Weighted NTF Synthesis for Cost-Efficient ΣΔ based DPD Feedback Loops
  
  • Zeng Haoyang
  • Pham Dang-Kièn Germain
  • Mohellebi Reda
  • Jabbour Chadi
  , 2026, pp.1-5. (10.1109/NewCAS64543.2026.11674034)
  DOI : 10.1109/NewCAS64543.2026.11674034
• Joint Visibility Analysis of RIS in Non-Terrestrial Networks through Stochastic Geometry
  
  • Balakrishnan Ashutosh
  • Lee Junse
  • Baccelli François
  , 2026. Non-Terrestrial networks (NTNs) are a key theme in upcoming 6G communications, especially for ubiquitous coverage. Urban environments, comprising of high rise buildings often result in blocking the line of sight (LoS) path between the user equipment (UE) and the NTN base station (NTN-BS). In this paper we investigate the situation where reconfigurable intelligent surfaces (RIS) are deployed on the building roof-tops to ensure multi-hop connectivity between the UE and the NTN-BS. In such a scenario, it becomes crucial to statistically study the LoS visibility of the RIS from the UE as well as from the NTN-BS, hence termed as joint visibility. In this work, accounting for the dual stochasticity arising from the locations of the RIS deployed buildings and the respective random building heights, we statistically study the probability of joint RIS visibility in a two-dimensional (2D) scenario considering a deterministic location of the NTN-BS. Further, we study the joint RIS visibility statistics conditional on the UE-NTN link being LoS or non-LoS. For the RISs deployed as a point point process (PPP) having exponentially distributed heights, the expected RISs jointly visible under the unconditional and conditional geometric settings are derived in closed form. Interestingly, in the 2D setting, the maximum expected RISs jointly visible, unconditionally, is twice the Basel number (π^2 /6). The simulated results are analyzed over building density, average building height, the altitude and position of the NTN-BS. We also illustrate probability heatmaps, demonstrating the strongest chance to have a RIS used conditioned on the system geometry. This study is expected to be useful in planning the deployment of RIS in urban areas, improving the signal and for assessing economic aspects.
• Transitions as the Native Objects of Dispersive Light-Matter Dynamics
  
  • Mohamed Meguebel
  • Federico Maxime
  • Garbe Louis
  • Belabas Nadia
  • Fabre Nicolas
  , 2026. We introduce a framework where light-matter transitions, rather than states, are the primary dynamical objects. Successive compositions of elementary transitions yield multiphoton processes with compact diagrammatic bookkeeping of resonant and off-resonant pathways. This approach enables transparent derivations of effective high-order Hamiltonians in the dispersive regime, foundational to quantum-information applications. Applied to the paradigmatic Jaynes-Cummings model, our framework reveals a photon-number-independent intrinsic Rabi frequency and persistent polaritonic hybridization in the dispersive regime, unifying resonant and dispersive limits.
• Conception de circuits de lecture à faible consommation énergétique ​pour capteurs d’image en technologie CMOS 18nm FD-SOI
  
  • Perrier Valentin
  • Mas Alexandre
  • Quémerais Thomas
  • Larras Benoit
  • Mohellebi Reda
  • Frappé Antoine
  • Desgreys Patricia
  , 2026.
• Time-frequency Talbot effect as Clifford operations on entangled time-frequency GKP states
  
  • Pousset Thomas
  • Dalidet Romain
  • Labonté Laurent
  • Fabre Nicolas
  , 2026. The Talbot effect-a near-field diffraction phenomenon in which a periodic wavefront self-images at regular distances-can be transposed to the time-frequency domain via the space-time duality between diffraction and dispersive broadening. We exploit this analogy to define the time-frequency (TF) Talbot effect and show that it implements different Clifford operations on TF Gottesman-Kitaev-Preskill (TF-GKP) qubits (Phys. Rev. A 102, 012607), a class of qubit states encoded in the discretised frequency and time-of-arrival degrees of freedom of entangled photon pairs, whose logical basis corresponds to even and odd components of an entangled frequency combs. These states are intrinsically robust against small frequency and temporal displacements, which can be further corrected by linear or nonlinear quantum error-correction schemes. We analyse the role of the comb envelope and peak width relative to the free spectral range, and show that a compromise must be made between the gate fidelity of the Clifford gates induced by TF-Talbot operation and the error-correction capacity of the code. We then demonstrate that the signature of the TF-Talbot effect is directly accessible via the generalised Hong-Ou-Mandel interferometer: all six logical GKP states can be unambiguously distinguished by introducing a frequency shift of half the comb periodicity in one interferometer arm. We conclude with a feasibility analysis based on current experimental technology, identifying the comb finesse as the key figure of merit for both gate performance and correctability. This conclusion extends naturally to quadrature GKP states, where a shear in quadrature phase space is precisely a Talbot effect. (10.1103/s8jp-l4hf)
  DOI : 10.1103/s8jp-l4hf
• Analytical solution of radiative transfer equation of light radiance in turbid slab with inner-medium source under P3-1D approximation
  
  • Fade Julien
  • Roux Aurélien
  • Breton Vicente Alberto
  • Sorrente Béatrice
  • Grillot Frédéric
  • Da Silva Anabela
  Optics Express, Optical Society of America - OSA Publishing, 2026, 34 (13), pp.23586. A generalized model of the 1-dimensional radiative transfer equation of the light radiance in a turbid slad is detailed, under the P-3 approximation, including the possibility to model a continuous plane-wave source located at any depth within the scattering slab. This analytical model, which requires significant evolution of the P3-1D model is extensively described and validated by comparison with Monte-Carlo numerical experiments. A series of numerical simulations illustrates some of the modelling possibilities offered by this extended model, which makes it possible to continuously model the transition between a classical slab geometry and a semi-infinite geometry. (10.1364/OE.599850)
  DOI : 10.1364/OE.599850
• Effective Hamiltonians in Cavity and Waveguide QED from Transition-Operator Diagrammatic Perturbation Theory
  
  • Meguebel Mohamed
  • Federico Maxime
  • Garbe Louis
  • Belabas Nadia
  • Fabre Nicolas
  , 2026. We propose an adiabatic-elimination formalism in the dispersive regime based on a transitioncentric perturbation theory. The perturbative expansion is recast into a diagrammatic framework, while adiabatic elimination is implemented through controlled projections onto transition subspaces. Our approach applies systematically at arbitrary perturbation order, and is suited to multilevel systems and multiple qubits in both cavity and waveguide quantum electrodynamics. It ultimately enables the explicit construction of effective higher-order Hamiltonians while bypassing important limitations of existing techniques, thereby providing a practical toolbox for multiphoton processes in the dispersive regime.
• Scalable Models and Simulations for Communicating Embedded Systems
  
  • Rayon Malou
  • Genius Daniela
  • Apvrille Ludovic
  , 2026.
• A Max–Min Neural Network Model for Propositional Fuzzy Logic
  
  • Baaj Ismaïl
  • Bloch Isabelle
  • Rico Agnès
  • Strauss Olivier
  , 2026, 3020, pp.334-348. We propose and study a neural network model for propositional fuzzy logic over $[0,1]$, based on a single max-min neuron. The neuron output is obtained by combining $\max$, $\min$, and a switch parameter through a fuzzy equivalence, and when the switch is Boolean, it reduces to a strict choice between the value obtained by max-min aggregation of the input and weights and its negation. We show that cascades of such neurons can represent the elementary Boolean connectives, and we analyze the neuron equation from two complementary viewpoints: solving for the parameters when the input and output are fixed, and solving for admissible inputs when the parameters and output are fixed. These results are then used to describe a training scheme by forward evaluation and backward propagation of constraints through the layers. (10.1007/978-3-032-28997-1_24)
  DOI : 10.1007/978-3-032-28997-1_24
• Methods for Sizing and Deploying Virtualised Mobile Networks : Application to 5G/6G, Optimisation for Smart Electrical Grids
  
  • Altawil Rosy
  , 2026. Modernising industrial smart grids operated by EDF requires highly reliable communications for Protection, Automation, and Control (PAC) systems. Although 5G network slicing supports mission-critical messages via URLLC, industrial clients lack visibility into the operator's core network, making it hard to guarantee the required 10−5 packet loss probability. Furthermore, traditional allocation models based on Poisson's law fail to capture real-world traffic burstiness, causing resource deficits 30% of the time. Tobridge this visibility gap, a Machine Learning classification model was developed. A LightGBM model was implemented to classify trafficinto URLLC and eMBB slices using experimental data restricted to the lower PHY and MAC layers. Achieving a 97.6% accuracy demonstrates that industrial clients can autonomously monitor traffic without operator-level KPIs. Building on this, a dynamic resource allocation model using Deep Reinforcement Learning (DRL) was introduced. A Deep Q-Network (DQN) agent continuously interacts with the network environment across a continuous state space (URLLC/eMBB loads, cell capacity) to manage the explorationexploitation trade-off. The agent dynamically allocates the optimal guard channels to minimize URLLC packet loss while maintaining fairness toward the eMBB buffer. Combining queuing limits, ML classification, and DQN, this thesis empowers industrial clients to autonomously predict, dimension, and adjust their 5G network slice resources instantaneously.
• CALICE: Continuous bitrate control with Adapted LIC modEl
  
  • Spadaro Gabriele
  • Presta Alberto
  • Giraldo Jhony
  • Fiandrotti Attilio
  • Grangetto Marco
  • Tartaglione Enzo
  ACM Transactions on Multimedia Computing, Communications and Applications, Association for Computing Machinery, 2026, pp.1-21. Learned image compression (LIC) has drawn much attention recently as it outperforms standardized codecs in rate-distortion (RD) efficiency. However, a LIC model is typically trained for a specific RD trade-off, and achieving a different target rate requires retraining the model and storing the weights as a whole, limiting the practical applicability of LIC. In this paper, we introduce CALICE, a framework for achieving continuous bitrate control by plugging into a pre-trained LIC model a set of modular adapters. Unlike similar methods that require a distinct set of adapters for each target rate, our method achieves continuous bitrate control by modulating a single set of adapters via a scalar parameter \(\boldsymbol{\alpha}\) , with a total overhead of less than \(\mathbf{0.35}\boldsymbol{\%}\) of the parameters of the LIC model. This design enables efficient support for multiple distortion objectives by learning lightweight, distortion-aware adapters. We also extend our strategy beyond rate control, demonstrating its ability to provide fine-grained adaptation of perceptual quality along the distortion-perception trade-off. To our knowledge, this is the first method that jointly addresses rate and perceptual control using a unified, low-cost strategy. We publicly released the code at https://github.com/EIDOSLAB/CALICE . (10.1145/3820658)
  DOI : 10.1145/3820658