Search

The increased technological complexity and demand for software reliability require organizations to formally design and verify their safety-critical programs to minimize systematic failures. Formal methods are recommended by functional safety standards (e.g., by IEC 61511 for the process industry and by the generic IEC 61508) and play a crucial role. Their structured approach reduces ambiguity in system requirements, facilitating early error detection. This paper introduces a formal verification service for PLC (programmable logic controller) programs compliant with functional safety standards, providing external expertise to organizations while eliminating the need for extensive internal training. It offers a cost-effective solution to meet the rising demands for formal verification processes. The approach is extended to include modeling time-dependent, know-how-protected components, enabling formal verification of real safety-critical applications. A case study shows the application of PLC formal verification as a service provided by CERN in a safety-critical installation at the GSI particle accelerator facility., Comment: Extended version of NFM 2025 paper

Myocardial perfusion cardiovascular magnetic resonance (pCMR) using first-pass contrast-enhanced imaging could play an important role in the detection of epicardial and microvascular coronary artery disease. Recently, the emergence of quantitative pCMR has provided a more reliable and observer-independent analysis compared to visual interpretation of dynamic images. This review aims to cover the basics of quantitative pCMR, from acquisition protocols, its use in preclinical and clinical studies, image reconstruction and motion handling, to automated quantitative pCMR pipelines. It also offers an overview of emerging tools in the field, including artificial intelligence-based methods., Comment: All copyright for this preprint belongs to the authors. The work is protected under copyright and may not be reproduced, distributed, or used without proper citation and permission, unless otherwise specified

We present LiGR, a large-scale ranking framework developed at LinkedIn that brings state-of-the-art transformer-based modeling architectures into production. We introduce a modified transformer architecture that incorporates learned normalization and simultaneous set-wise attention to user history and ranked items. This architecture enables several breakthrough achievements, including: (1) the deprecation of most manually designed feature engineering, outperforming the prior state-of-the-art system using only few features (compared to hundreds in the baseline), (2) validation of the scaling law for ranking systems, showing improved performance with larger models, more training data, and longer context sequences, and (3) simultaneous joint scoring of items in a set-wise manner, leading to automated improvements in diversity. To enable efficient serving of large ranking models, we describe techniques to scale inference effectively using single-pass processing of user history and set-wise attention. We also summarize key insights from various ablation studies and A/B tests, highlighting the most impactful technical approaches.

This paper studies the Bayesian regret of the Thompson Sampling algorithm for bandit problems, building on the information-theoretic framework introduced by Russo and Van Roy (2015). Specifically, it extends the rate-distortion analysis of Dong and Van Roy (2018), which provides near-optimal bounds for linear bandits. A limitation of these results is the assumption of a finite action space. We address this by extending the analysis to settings with infinite and continuous action spaces. Additionally, we specialize our results to bandit problems with expected rewards that are Lipschitz continuous with respect to the action space, deriving a regret bound that explicitly accounts for the complexity of the action space., Comment: 5 pages, accepted to ICASSP

Text datasets can be represented using models that do not preserve text structure, or using models that preserve text structure. Our hypothesis is that depending on the dataset nature, there can be advantages using a model that preserves text structure over one that does not, and viceversa. The key is to determine the best way of representing a particular dataset, based on the dataset itself. In this work, we propose to investigate this problem by combining text distortion and algorithmic clustering based on string compression. Specifically, a distortion technique previously developed by the authors is applied to destroy text structure progressively. Following this, a clustering algorithm based on string compression is used to analyze the effects of the distortion on the information contained in the texts. Several experiments are carried out on text datasets and artificially-generated datasets. The results show that in strongly structural datasets the clustering results worsen as text structure is progressively destroyed. Besides, they show that using a compressor which enables the choice of the size of the left-context symbols helps to determine the nature of the datasets. Finally, the results are contrasted with a method based on multidimensional projections and analogous conclusions are obtained.

Scaling laws have emerged as important components of large language model (LLM) training as they can predict performance gains through scale, and provide guidance on important hyper-parameter choices that would otherwise be expensive. LLMs also rely on large, high-quality training datasets, like those sourced from (sometimes sensitive) user data. Training models on this sensitive user data requires careful privacy protections like differential privacy (DP). However, the dynamics of DP training are significantly different, and consequently their scaling laws are not yet fully understood. In this work, we establish scaling laws that accurately model the intricacies of DP LLM training, providing a complete picture of the compute-privacy-utility tradeoffs and the optimal training configurations in many settings.

We exhibit a uniform method for obtaining (wellfounded and non-wellfounded) cut-free sequent-style proof systems that are sound and complete for various classes of action algebras, i.e., Kleene algebras enriched with meets and residuals. Our method applies to any class of *-continuous action algebras that is defined, relative to the class of all *-continuous action algebras, by analytic quasiequations. The latter make up an expansive class of conditions encompassing the algebraic analogues of most well-known structural rules. These results are achieved by wedding existing work on non-wellfounded proof theory for action algebras with tools from algebraic proof theory.

Time series anomaly detection presents various challenges due to the sequential and dynamic nature of time-dependent data. Traditional unsupervised methods frequently encounter difficulties in generalization, often overfitting to known normal patterns observed during training and struggling to adapt to unseen normality. In response to this limitation, self-supervised techniques for time series have garnered attention as a potential solution to undertake this obstacle and enhance the performance of anomaly detectors. This paper presents a comprehensive review of the recent methods that make use of self-supervised learning for time series anomaly detection. A taxonomy is proposed to categorize these methods based on their primary characteristics, facilitating a clear understanding of their diversity within this field. The information contained in this survey, along with additional details that will be periodically updated, is available on the following GitHub repository: https://github.com/Aitorzan3/Awesome-Self-Supervised-Time-Series-Anomaly-Detection.

Euclidean AdS wormholes provide a natural setup for studying the AdS/CFT correspondence with multiple boundaries. However, from a bottom-up perspective, they cannot be embedded in the four-dimensional Einstein-AdS-Maxwell theory if these boundaries have positive curvature. Nevertheless, Maldacena and Maoz showed that this obstruction could be circumvented by introducing merons in the four-dimensional Einstein-AdS-Yang-Mills theory. In this work, we show that Euclidean-AdS wormholes also exist in the four-dimensional Einstein-AdS-Skyrme theory, whose matter sector possesses a nontrivial baryonic charge. We compute its free energy and show that it does not depend on the integration constants whatsoever, resembling topological solitons. Additionally, we obtain its holographic stress tensor and show that it vanishes, allowing us to interpret this configuration as a holographic Bogomol'nyi-Prasad-Sommerfield (BPS) state. Other topologically nontrivial ground states in Einstein-Skyrme theory are found, such as gravitational instantons, representing the homotopically inequivalent vacua of the theory. We find that they develop Hawking-Page phase transitions above a critical temperature. Some of these solutions are periodic in Euclidean time, representing the gravitational analog of calorons in Yang-Mills theory., Comment: 22 pages

We often interact with untrusted parties. Prioritization of privacy can limit the effectiveness of these interactions, as achieving certain goals necessitates sharing private data. Traditionally, addressing this challenge has involved either seeking trusted intermediaries or constructing cryptographic protocols that restrict how much data is revealed, such as multi-party computations or zero-knowledge proofs. While significant advances have been made in scaling cryptographic approaches, they remain limited in terms of the size and complexity of applications they can be used for. In this paper, we argue that capable machine learning models can fulfill the role of a trusted third party, thus enabling secure computations for applications that were previously infeasible. In particular, we describe Trusted Capable Model Environments (TCMEs) as an alternative approach for scaling secure computation, where capable machine learning model(s) interact under input/output constraints, with explicit information flow control and explicit statelessness. This approach aims to achieve a balance between privacy and computational efficiency, enabling private inference where classical cryptographic solutions are currently infeasible. We describe a number of use cases that are enabled by TCME, and show that even some simple classic cryptographic problems can already be solved with TCME. Finally, we outline current limitations and discuss the path forward in implementing them.

Periodic dynamical systems, distinguished by their repetitive behavior over time, are prevalent across various engineering disciplines. In numerous applications, particularly within industrial contexts, the implementation of model predictive control (MPC) schemes tailored to optimize specific economic criteria was shown to offer substantial advantages. However, the real-time implementation of these schemes is often infeasible due to limited computational resources. To tackle this problem, we propose a resource-efficient economic model predictive control scheme for periodic systems, leveraging existing single-layer MPC techniques. Our method relies on a single quadratic optimization problem, which ensures high computational efficiency for real-time control in dynamic settings. We prove feasibility, stability and convergence to optimum of the proposed approach, and validate the effectiveness through numerical experiments.

Collisionless cross-field electron transport in an E x B configuration relevant for electric propulsion is studied using data from a (z, {\theta}) full-PIC simulation. Higher-order spectral analysis shows that transport is dominated by the in-phase interaction of the oscillations of the azimuthal electric field and the electron density associated to the first electron cyclotron drift instability (ECDI) mode. A secondary contribution emanates from a lower-frequency mode, not predicted by linear ECDI theory, while higher modes have a minor direct impact on transport. However, a bicoherence analysis reveals that strong phase couplings exist among the ECDI modes, and a sparse symbolic regression spectral model, based on the three-wave coupling equations, suggests an inverse energy cascade as the most likely explanation, thus suggesting that higher modes contribute indirectly to transport by quadratic power transfer to the first mode. This work provides new insights into the dynamics of anomalous plasma transport in E x B sources and the underlying processes governing energy distribution across different scales, and supports the validity of weak turbulence theory to examine their behavior.

Background: The worst outcomes of diabetic retinopathy (DR) can be prevented by implementing DR screening programs assisted by AI. At the University Hospital of Navarre (HUN), Spain, general practitioners (GPs) grade fundus images in an ongoing DR screening program, referring to a second screening level (ophthalmologist) target patients. Methods: After collecting their requirements, HUN decided to develop a custom AI tool, called NaIA-RD, to assist their GPs in DR screening. This paper introduces NaIA-RD, details its implementation, and highlights its unique combination of DR and retinal image quality grading in a single system. Its impact is measured in an unprecedented before-and-after study that compares 19,828 patients screened before NaIA-RD's implementation and 22,962 patients screened after. Results: NaIA-RD influenced the screening criteria of 3/4 GPs, increasing their sensitivity. Agreement between NaIA-RD and the GPs was high for non-referral proposals (94.6% or more), but lower and variable (from 23.4\% to 86.6%) for referral proposals. An ophthalmologist discarded a NaIA-RD error in most of contradicted referral proposals by labeling the 93% of a sample of them as referable. In an autonomous setup, NaIA-RD would have reduced the study visualization workload by 4.27 times without missing a single case of sight-threatening DR referred by a GP. Conclusion: DR screening was more effective when supported by NaIA-RD, which could be safely used to autonomously perform the first level of screening. This shows how AI devices, when seamlessly integrated into clinical workflows, can help improve clinical pathways in the long term.

We investigate a terahertz (THz) gas sensing platform based on all-dielectric metasurfaces that support quasi-bound states in the continuum (quasi-BIC) with both electric and magnetic dipole resonances. The structure is designed to achieve the first Kerker condition, minimizing backscattering and maximizing light-matter interaction, which significantly enhances the sensitivity of the sensor. By optimizing structural parameters, this metasurface selectively resonates at characteristic absorption frequencies of target gases, facilitating detection even at low concentrations. We validate the approach using two gases with strong but distinct THz absorption profiles: hydrogen cyanide (HCN) and sulfur dioxide (SO2). Furthermore, the free-standing design maximizes gas interaction on both sides of the metasurface, eliminating substrate-induced losses and enabling a reduced physical footprint. Our findings indicate that this metasurface outperforms standard THz sensing approaches in terms of compactness and sensitivity per path length unit, obtaining the same detection threshold as sensing in free space with a path length between 2 and 3 orders of magnitude shorter, underscoring its potential for industrial applications where the available space for sensing can be limited., Comment: 12 pages, 5 figures

Federated learning is a versatile framework for training models in decentralized environments. However, the trust placed in clients makes federated learning vulnerable to backdoor attacks launched by malicious participants. While many defenses have been proposed, they often fail short when facing heterogeneous data distributions among participating clients. In this paper, we propose a novel defense mechanism for federated learning systems designed to mitigate backdoor attacks on the clients-side. Our approach leverages adversarial learning techniques and model patching to neutralize the impact of backdoor attacks. Through extensive experiments on the MNIST and Fashion-MNIST datasets, we demonstrate that our defense effectively reduces backdoor accuracy, outperforming existing state-of-the-art defenses, such as LFighter, FLAME, and RoseAgg, in i.i.d. and non-i.i.d. scenarios, while maintaining competitive or superior accuracy on clean data.

This paper introduces BIMCaP, a novel method to integrate mobile 3D sparse LiDAR data and camera measurements with pre-existing building information models (BIMs), enhancing fast and accurate indoor mapping with affordable sensors. BIMCaP refines sensor poses by leveraging a 3D BIM and employing a bundle adjustment technique to align real-world measurements with the model. Experiments using real-world open-access data show that BIMCaP achieves superior accuracy, reducing translational error by over 4 cm compared to current state-of-the-art methods. This advancement enhances the accuracy and cost-effectiveness of 3D mapping methodologies like SLAM. BIMCaP's improvements benefit various fields, including construction site management and emergency response, by providing up-to-date, aligned digital maps for better decision-making and productivity. Link to the repository: https://github.com/MigVega/BIMCaP, Comment: 10 pages, 24 figures, Conference: EG-ICE: 31st International Workshop on Intelligent Computing in Engineering

We study the performance of the Thompson Sampling algorithm for logistic bandit problems. In this setting, an agent receives binary rewards with probabilities determined by a logistic function, $\exp(\beta \langle a, \theta \rangle)/(1+\exp(\beta \langle a, \theta \rangle))$, with slope parameter $\beta>0$, and where both the action $a\in \mathcal{A}$ and parameter $\theta \in \mathcal{O}$ lie within the $d$-dimensional unit ball. Adopting the information-theoretic framework introduced by Russo and Van Roy (2016), we analyze the information ratio, a statistic that quantifies the trade-off between the immediate regret incurred and the information gained about the optimal action. We improve upon previous results by establishing that the information ratio is bounded by $\tfrac{9}{2}d\alpha^{-2}$, where $\alpha$ is a minimax measure of the alignment between the action space $\mathcal{A}$ and the parameter space $\mathcal{O}$, and is independent of $\beta$. Using this result, we derive a bound of order $O(d/\alpha\sqrt{T \log(\beta T/d)})$ on the Bayesian expected regret of Thompson Sampling incurred after $T$ time steps. To our knowledge, this is the first regret bound for logistic bandits that depends only logarithmically on $\beta$ while being independent of the number of actions. In particular, when the action space contains the parameter space, the bound on the expected regret is of order $\tilde{O}(d \sqrt{T})$., Comment: 21 pages, under review

Recent years have seen a significant surge in complex AI systems for competitive programming, capable of performing at admirable levels against human competitors. While steady progress has been made, the highest percentiles still remain out of reach for these methods on standard competition platforms such as Codeforces. Here we instead focus on combinatorial competitive programming, where the target is to find as-good-as-possible solutions to otherwise computationally intractable problems, over specific given inputs. We hypothesise that this scenario offers a unique testbed for human-AI synergy, as human programmers can write a backbone of a heuristic solution, after which AI can be used to optimise the scoring function used by the heuristic. We deploy our approach on previous iterations of Hash Code, a global team programming competition inspired by NP-hard software engineering problems at Google, and we leverage FunSearch to evolve our scoring functions. Our evolved solutions significantly improve the attained scores from their baseline, successfully breaking into the top percentile on all previous Hash Code online qualification rounds, and outperforming the top human teams on several. Our method is also performant on an optimisation problem that featured in a recent held-out AtCoder contest., Comment: Technical report. 18 pages, 8 figures

The chemical enrichment in low-ionization nuclear emission-line regions (LINERs) is still an issue with spatial resolution spectroscopic data due to the lack of studies and the uncertainties in the nature of their ionizing source, despite being the most abundant type of active galaxies in the nearby Universe. Considering different scenarios for the ionizing source (hot old stellar populations, active galactic nuclei (AGN) or inefficient accretion disks), we analyze the implications of these assumptions to constrain the chemical content of the gas-phase interstellar medium (ISM). We used a sample of 105 galaxies from Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey, whose nuclear central spaxels show LINER-like emission. For each scenario considered, we built a grid of photoionization models (4928 models for each considered ionizing source) which are later used in the open-source code HII-CHI-Mistry, allowing us to estimate chemical abundance ratios such as 12+log(O/H) or log(N/O) and constrain the ionization parameter that characterize the ionized ISM in those galaxies. We obtain that oxygen abundances in the nuclear region of LINER-like galaxies spread over a wide range 8.08 < 12+log(O/H) < 8.89, with a median solar value (in agreement with previous studies) if AGN models are considered. Nevertheless, the derived nitrogen-to-oxygen ratio is much less affected by the assumptions on the ionizing source, and point towards suprasolar values (log(N/O) = -0.69). By comparing the different analyzed scenarios, we show that if hot old stellar populations were responsible of the ionization of the ISM a complex picture (such as outflows and/or inflows scaling with galaxy chemical abundance) would be needed to explain the chemical enrichment history, whereas the assumption of AGN activity is compatible with the standard scenario found in most galaxies., Comment: Revised version after language edition. 24 pages, 13 figures, 9 tables (5 as supplemented material). Accepted for publication in A&A

The offline reinforcement learning (RL) problem aims to learn an optimal policy from historical data collected by one or more behavioural policies (experts) by interacting with an environment. However, the individual experts may be privacy-sensitive in that the learnt policy may retain information about their precise choices. In some domains like personalized retrieval, advertising and healthcare, the expert choices are considered sensitive data. To provably protect the privacy of such experts, we propose a novel consensus-based expert-level differentially private offline RL training approach compatible with any existing offline RL algorithm. We prove rigorous differential privacy guarantees, while maintaining strong empirical performance. Unlike existing work in differentially private RL, we supplement the theory with proof-of-concept experiments on classic RL environments featuring large continuous state spaces, demonstrating substantial improvements over a natural baseline across multiple tasks.

Differentially Private Stochastic Gradient Descent (DP-SGD) is a popular method for training machine learning models with formal Differential Privacy (DP) guarantees. As DP-SGD processes the training data in batches, it uses Poisson sub-sampling to select batches at each step. However, due to computational and compatibility benefits, replacing sub-sampling with shuffling has become common practice. Yet, since tight theoretical guarantees for shuffling are currently unknown, prior work using shuffling reports DP guarantees as though Poisson sub-sampling was used. This prompts the need to verify whether this discrepancy is reflected in a gap between the theoretical guarantees from state-of-the-art models and the actual privacy leakage. To do so, we introduce a novel DP auditing procedure to analyze DP-SGD with shuffling. We show that state-of-the-art DP models trained with shuffling appreciably overestimated privacy guarantees (up to 4x). In the process, we assess the impact of several parameters, such as batch size, privacy budget, and threat model, on privacy leakage. Finally, we study two variations of the shuffling procedure found in the wild, which result in further privacy leakage. Overall, our work empirically attests to the risk of using shuffling instead of Poisson sub-sampling vis-\`a-vis the actual privacy leakage of DP-SGD.

Content moderation typically combines the efforts of human moderators and machine learning models. However, these systems often rely on data where significant disagreement occurs during moderation, reflecting the subjective nature of toxicity perception. Rather than dismissing this disagreement as noise, we interpret it as a valuable signal that highlights the inherent ambiguity of the content,an insight missed when only the majority label is considered. In this work, we introduce a novel content moderation framework that emphasizes the importance of capturing annotation disagreement. Our approach uses multitask learning, where toxicity classification serves as the primary task and annotation disagreement is addressed as an auxiliary task. Additionally, we leverage uncertainty estimation techniques, specifically Conformal Prediction, to account for both the ambiguity in comment annotations and the model's inherent uncertainty in predicting toxicity and disagreement.The framework also allows moderators to adjust thresholds for annotation disagreement, offering flexibility in determining when ambiguity should trigger a review. We demonstrate that our joint approach enhances model performance, calibration, and uncertainty estimation, while offering greater parameter efficiency and improving the review process in comparison to single-task methods., Comment: 35 pages, 1 figure

Interoperability is the key to implementing a wide range of energy systems applications. It involves the seamless cooperation of different methods and components. With smart energy systems, interoperability faces challenges due to integrating differ-ent approaches and technologies. This includes dealing with heterogeneous approaches with various communication proto-cols and data formats. However, it is essential for smart energy systems to carry out thorough interoperability tests. They are usually diverse, and challenging, thus requiring careful consideration of compatibility issues and complex integration scenari-os. Overcoming these challenges requires a systematic approach that includes thorough test planning, rigorous testing, and continuous test monitoring. Although numerous testing approaches exist, most are more developed at the component/device level than at the system level. Consequently, there are few approaches and related facilities to test the interoperability of smart energy approaches and solutions at the system level. This work analyses existing interoperability test concepts, identi-fies enablers and the potential for harmonisation of procedures, and proposes further developments of these approaches., Comment: 14th Mediterranean Conference on Power Generation Transmission, Distribution and Energy Conversion (MED POWER 2024)

We propose a method to enhance the performance of Large Language Models (LLMs) by integrating quantum computing and quantum-inspired techniques. Specifically, our approach involves replacing the weight matrices in the Self-Attention and Multi-layer Perceptron layers with a combination of two variational quantum circuits and a quantum-inspired tensor network, such as a Matrix Product Operator (MPO). This substitution enables the reproduction of classical LLM functionality by decomposing weight matrices through the application of tensor network disentanglers and MPOs, leveraging well-established tensor network techniques. By incorporating more complex and deeper quantum circuits, along with increasing the bond dimensions of the MPOs, our method captures additional correlations within the quantum-enhanced LLM, leading to improved accuracy beyond classical models while maintaining low memory overhead., Comment: 4 pages, 2 figures

We study agents acting in an unknown environment where the agent's goal is to find a robust policy. We consider robust policies as policies that achieve high cumulative rewards for all possible environments. To this end, we consider agents minimizing the maximum regret over different environment parameters, leading to the study of minimax regret. This research focuses on deriving information-theoretic bounds for minimax regret in Markov Decision Processes (MDPs) with a finite time horizon. Building on concepts from supervised learning, such as minimum excess risk (MER) and minimax excess risk, we use recent bounds on the Bayesian regret to derive minimax regret bounds. Specifically, we establish minimax theorems and use bounds on the Bayesian regret to perform minimax regret analysis using these minimax theorems. Our contributions include defining a suitable minimax regret in the context of MDPs, finding information-theoretic bounds for it, and applying these bounds in various scenarios.

We propose a simple heuristic privacy analysis of noisy clipped stochastic gradient descent (DP-SGD) in the setting where only the last iterate is released and the intermediate iterates remain hidden. Namely, our heuristic assumes a linear structure for the model. We show experimentally that our heuristic is predictive of the outcome of privacy auditing applied to various training procedures. Thus it can be used prior to training as a rough estimate of the final privacy leakage. We also probe the limitations of our heuristic by providing some artificial counterexamples where it underestimates the privacy leakage. The standard composition-based privacy analysis of DP-SGD effectively assumes that the adversary has access to all intermediate iterates, which is often unrealistic. However, this analysis remains the state of the art in practice. While our heuristic does not replace a rigorous privacy analysis, it illustrates the large gap between the best theoretical upper bounds and the privacy auditing lower bounds and sets a target for further work to improve the theoretical privacy analyses. We also empirically support our heuristic and show existing privacy auditing attacks are bounded by our heuristic analysis in both vision and language tasks., Comment: ICLR 2025 camera-ready version

In an era where maritime infrastructures are crucial, advanced situational awareness solutions are increasingly important. The use of optical camera systems can allow real-time usage of maritime footage. This thesis presents an investigation into leveraging deep learning and computer vision to advance real-time ship recognition and georeferencing for the improvement of maritime situational awareness. A novel dataset, ShipSG, is introduced, containing 3,505 images and 11,625 ship masks with corresponding class and geographic position. After an exploration of state-of-the-art, a custom real-time segmentation architecture, ScatYOLOv8+CBAM, is designed for the NVIDIA Jetson AGX Xavier embedded system. This architecture adds the 2D scattering transform and attention mechanisms to YOLOv8, achieving an mAP of 75.46% and an 25.3 ms per frame, outperforming state-of-the-art methods by over 5%. To improve small and distant ship recognition in high-resolution images on embedded systems, an enhanced slicing mechanism is introduced, improving mAP by 8% to 11%. Additionally, a georeferencing method is proposed, achieving positioning errors of 18 m for ships up to 400 m away and 44 m for ships between 400 m and 1200 m. The findings are also applied in real-world scenarios, such as the detection of abnormal ship behaviour, camera integrity assessment and 3D reconstruction. The approach of this thesis outperforms existing methods and provides a framework for integrating recognized and georeferenced ships into real-time systems, enhancing operational effectiveness and decision-making for maritime stakeholders. This thesis contributes to the maritime computer vision field by establishing a benchmark for ship segmentation and georeferencing research, demonstrating the viability of deep-learning-based recognition and georeferencing methods for real-time maritime monitoring.

Differential privacy (DP) offers a robust framework for safeguarding individual data privacy. To utilize DP in training modern machine learning models, differentially private optimizers have been widely used in recent years. A popular approach to privatize an optimizer is to clip the individual gradients and add sufficiently large noise to the clipped gradient. This approach led to the development of DP optimizers that have comparable performance with their non-private counterparts in fine-tuning tasks or in tasks with a small number of training parameters. However, a significant performance drop is observed when these optimizers are applied to large-scale training. This degradation stems from the substantial noise injection required to maintain DP, which disrupts the optimizer's dynamics. This paper introduces DiSK, a novel framework designed to significantly enhance the performance of DP optimizers. DiSK employs Kalman filtering, a technique drawn from control and signal processing, to effectively denoise privatized gradients and generate progressively refined gradient estimations. To ensure practicality for large-scale training, we simplify the Kalman filtering process, minimizing its memory and computational demands. We establish theoretical privacy-utility trade-off guarantees for DiSK, and demonstrate provable improvements over standard DP optimizers like DPSGD in terms of iteration complexity upper-bound. Extensive experiments across diverse tasks, including vision tasks such as CIFAR-100 and ImageNet-1k and language fine-tuning tasks such as GLUE, E2E, and DART, validate the effectiveness of DiSK. The results showcase its ability to significantly improve the performance of DP optimizers, surpassing state-of-the-art results under the same privacy constraints on several benchmarks.

Linguistic fieldwork is an important component in language documentation and preservation. However, it is a long, exhaustive, and time-consuming process. This paper presents a novel model that guides a linguist during the fieldwork and accounts for the dynamics of linguist-speaker interactions. We introduce a novel framework that evaluates the efficiency of various sampling strategies for obtaining morphological data and assesses the effectiveness of state-of-the-art neural models in generalising morphological structures. Our experiments highlight two key strategies for improving the efficiency: (1) increasing the diversity of annotated data by uniform sampling among the cells of the paradigm tables, and (2) using model confidence as a guide to enhance positive interaction by providing reliable predictions during annotation.

Advances in generative AI point towards a new era of personalized applications that perform diverse tasks on behalf of users. While general AI assistants have yet to fully emerge, their potential to share personal data raises significant privacy challenges. This paper introduces CI-Bench, a comprehensive synthetic benchmark for evaluating the ability of AI assistants to protect personal information during model inference. Leveraging the Contextual Integrity framework, our benchmark enables systematic assessment of information flow across important context dimensions, including roles, information types, and transmission principles. We present a novel, scalable, multi-step synthetic data pipeline for generating natural communications, including dialogues and emails. Unlike previous work with smaller, narrowly focused evaluations, we present a novel, scalable, multi-step data pipeline that synthetically generates natural communications, including dialogues and emails, which we use to generate 44 thousand test samples across eight domains. Additionally, we formulate and evaluate a naive AI assistant to demonstrate the need for further study and careful training towards personal assistant tasks. We envision CI-Bench as a valuable tool for guiding future language model development, deployment, system design, and dataset construction, ultimately contributing to the development of AI assistants that align with users' privacy expectations.

Data linkage is increasingly used in health research and policy making and is relied on for understanding health inequalities. However, linked data is only as useful as the underlying data quality, and differential linkage rates may induce selection bias in the linked data. A mechanism that selectively compromises data quality is name romanisation. Converting text of a different writing system into Latin based writing, or romanisation, has long been the standard process of representing names in character-based writing systems such as Chinese, Vietnamese, and other languages such as Swahili. Unstandardised romanisation of Chinese characters, due in part to problems of preserving the correct name orders the lack of proper phonetic representation of a tonal language, has resulted in poor linkage rates for Chinese immigrants. This opinion piece aims to suggests that the use of standardised romanisation systems for Cantonese (Jyutping) or Mandarin (Pinyin) Chinese, which incorporate tonal information, could improve linkage rates and accuracy for individuals with Chinese names. We used 771 Chinese and English names scraped from openly available sources, and compared the utility of Jyutping, Pinyin and the Hong Kong Government Romanisation system (HKG-romanisation) for representing Chinese names. We demonstrate that both Jyutping and Pinyin result in fewer errors compared with the HKG-romanisation system. We suggest that collecting and preserving people's names in their original writing systems is ethically and socially pertinent. This may inform development of language-specific pre-processing and linkage paradigms that result in more inclusive research data which better represents the targeted populations.

Here we introduce the application of Tensor Networks (TN) to launch attacks on symmetric-key cryptography. Our approaches make use of Matrix Product States (MPS) as well as our recently-introduced Flexible-PEPS Quantum Circuit Simulator (FQCS). We compare these approaches with traditional brute-force attacks and Variational Quantum Attack Algorithm (VQAA) methods also proposed by us. Our benchmarks include the Simplified Data Encryption Standard (S-DES) with 10-bit keys, Simplified Advanced Encryption Standard (S-AES) with 16-bit keys, and Blowfish with 32-bit keys. We find that for small key size, MPS outperforms VQAA and FQCS in both time and average iterations required to recover the key. As key size increases, FQCS becomes more efficient in terms of average iterations compared to VQAA and MPS, while MPS remains the fastest in terms of time. These results highlight the potential of TN methods in advancing quantum cryptanalysis, particularly in optimizing both speed and efficiency. Our results also show that entanglement becomes crucial as key size increases., Comment: 10 pages, 8 figures

We introduce a novel method for updating 3D geospatial models, specifically targeting occlusion removal in large-scale maritime environments. Traditional 3D reconstruction techniques often face problems with dynamic objects, like cars or vessels, that obscure the true environment, leading to inaccurate models or requiring extensive manual editing. Our approach leverages deep learning techniques, including instance segmentation and generative inpainting, to directly modify both the texture and geometry of 3D meshes without the need for costly reprocessing. By selectively targeting occluding objects and preserving static elements, the method enhances both geometric and visual accuracy. This approach not only preserves structural and textural details of map data but also maintains compatibility with current geospatial standards, ensuring robust performance across diverse datasets. The results demonstrate significant improvements in 3D model fidelity, making this method highly applicable for maritime situational awareness and the dynamic display of auxiliary information., Comment: Preprint of SPIE Sensor + Imaging 2024 conference paper

Visible light communication (VLC) complements radio frequency in indoor environments with large wireless data traffic. However, VLC is hindered by dramatic path losses when an opaque object is interposed between the transmitter and the receiver. Prior works propose the use of plane mirrors as optical reconfigurable intelligent surfaces (ORISs) to enhance communications through non-line-of-sight links. Plane mirrors rely on their orientation to forward the light to the target user location, which is challenging to implement in practice. This paper studies the potential of curved mirrors as static reflective surfaces to provide a broadening specular reflection that increases the signal coverage in mirror-assisted VLC scenarios. We study the behavior of paraboloid and semi-spherical mirrors and derive the irradiance equations. We provide extensive numerical and analytical results and show that curved mirrors, when developed with proper dimensions, may reduce the shadowing probability to zero, while static plane mirrors of the same size have shadowing probabilities larger than 65%. Furthermore, the signal-to-noise ratio offered by curved mirrors may suffice to provide connectivity to users deployed in the room even when a line-of-sight link blockage occurs., Comment: Accepted to be published in IEEE Globecom 2024

Finite element modeling is a well-established tool for structural analysis, yet modeling complex structures often requires extensive pre-processing, significant analysis effort, and considerable time. This study addresses this challenge by introducing an innovative method for real-time prediction of structural static responses using DeepOnet which relies on a novel approach to physics-informed networks driven by structural balance laws. This approach offers the flexibility to accurately predict responses under various load classes and magnitudes. The trained DeepONet can generate solutions for the entire domain, within a fraction of a second. This capability effectively eliminates the need for extensive remodeling and analysis typically required for each new case in FE modeling. We apply the proposed method to two structures: a simple 2D beam structure and a comprehensive 3D model of a real bridge. To predict multiple variables with DeepONet, we utilize two strategies: a split branch/trunk and multiple DeepONets combined into a single DeepONet. In addition to data-driven training, we introduce a novel physics-informed training approaches. This method leverages structural stiffness matrices to enforce fundamental equilibrium and energy conservation principles, resulting in two novel physics-informed loss functions: energy conservation and static equilibrium using the Schur complement. We use various combinations of loss functions to achieve an error rate of less than 5% with significantly reduced training time. This study shows that DeepONet, enhanced with hybrid loss functions, can accurately and efficiently predict displacements and rotations at each mesh point, with reduced training time.

Federated Learning (FL) is an innovative approach to distributed machine learning. While FL offers significant privacy advantages, it also faces security challenges, particularly from poisoning attacks where adversaries deliberately manipulate local model updates to degrade model performance or introduce hidden backdoors. Existing defenses against these attacks have been shown to be effective when the data on the nodes is identically and independently distributed (i.i.d.), but they often fail under less restrictive, non-i.i.d data conditions. To overcome these limitations, we introduce Celtibero, a novel defense mechanism that integrates layered aggregation to enhance robustness against adversarial manipulation. Through extensive experiments on the MNIST and IMDB datasets, we demonstrate that Celtibero consistently achieves high main task accuracy (MTA) while maintaining minimal attack success rates (ASR) across a range of untargeted and targeted poisoning attacks. Our results highlight the superiority of Celtibero over existing defenses such as FL-Defender, LFighter, and FLAME, establishing it as a highly effective solution for securing federated learning systems against sophisticated poisoning attacks.

In this article, we study the fundamental limits in the design of fair and/or private representations achieving perfect demographic parity and/or perfect privacy through the lens of information theory. More precisely, given some useful data $X$ that we wish to employ to solve a task $T$, we consider the design of a representation $Y$ that has no information of some sensitive attribute or secret $S$, that is, such that $I(Y;S) = 0$. We consider two scenarios. First, we consider a design desiderata where we want to maximize the information $I(Y;T)$ that the representation contains about the task, while constraining the level of compression (or encoding rate), that is, ensuring that $I(Y;X) \leq r$. Second, inspired by the Conditional Fairness Bottleneck problem, we consider a design desiderata where we want to maximize the information $I(Y;T|S)$ that the representation contains about the task which is not shared by the sensitive attribute or secret, while constraining the amount of irrelevant information, that is, ensuring that $I(Y;X|T,S) \leq r$. In both cases, we employ extended versions of the Functional Representation Lemma and the Strong Functional Representation Lemma and study the tightness of the obtained bounds. Every result here can also be interpreted as a coding with perfect privacy problem by considering the sensitive attribute as a secret.

We investigate the in-distribution generalization of machine learning algorithms. We depart from traditional complexity-based approaches by analyzing information-theoretic bounds that quantify the dependence between a learning algorithm and the training data. We consider two categories of generalization guarantees: 1) Guarantees in expectation: These bounds measure performance in the average case. Here, the dependence between the algorithm and the data is often captured by information measures. While these measures offer an intuitive interpretation, they overlook the geometry of the algorithm's hypothesis class. Here, we introduce bounds using the Wasserstein distance to incorporate geometry, and a structured, systematic method to derive bounds capturing the dependence between the algorithm and an individual datum, and between the algorithm and subsets of the training data. 2) PAC-Bayesian guarantees: These bounds measure the performance level with high probability. Here, the dependence between the algorithm and the data is often measured by the relative entropy. We establish connections between the Seeger--Langford and Catoni's bounds, revealing that the former is optimized by the Gibbs posterior. We introduce novel, tighter bounds for various types of loss functions. To achieve this, we introduce a new technique to optimize parameters in probabilistic statements. To study the limitations of these approaches, we present a counter-example where most of the information-theoretic bounds fail while traditional approaches do not. Finally, we explore the relationship between privacy and generalization. We show that algorithms with a bounded maximal leakage generalize. For discrete data, we derive new bounds for differentially private algorithms that guarantee generalization even with a constant privacy parameter, which is in contrast to previous bounds in the literature.

The available tools for damage identification in civil engineering structures are known to be computationally expensive and data-demanding. This paper proposes a comprehensive machine learning based damage identification (CMLDI) method that integrates modal analysis and dynamic analysis strategies. The proposed approach is applied to a real structure - KW51 railway bridge in Leuven. CMLDI diligently combines signal processing, machine learning (ML), and structural analysis techniques to achieve a fast damage identification solver that relies on minimal monitoring data. CMLDI considers modal analysis inputs and extracted features from acceleration responses to inform the damage identification based on the long-term and short-term monitoring data. Results of operational modal analysis, through the analysis of long-term monitoring data, are analyzed using pre-trained k-nearest neighbor (kNN) classifiers to identify damage existence, location, and magnitude. A well-crafted assembly of signal processing and ML methods is used to analyze acceleration time histories. Stacked gated recurrent unit (Stacked GRU) networks are used to identify damage existence, kNN classifiers are used to identify damage magnitude, and convolutions neural networks (CNN) are used to identify damage location. The damage identification results for the KW51 bridge demonstrate this approach's high accuracy, efficiency, and robustness. In this work, the training data is retrieved from the sensor of the KW51 bridge as well as the numerical finite element model (FEM). The proposed approach presents a systematic path to the generation of training data using a validated FEM. The data generation relies on modeling combinations of damage locations and magnitudes along the bridge.

Advanced AI assistants combine frontier LLMs and tool access to autonomously perform complex tasks on behalf of users. While the helpfulness of such assistants can increase dramatically with access to user information including emails and documents, this raises privacy concerns about assistants sharing inappropriate information with third parties without user supervision. To steer information-sharing assistants to behave in accordance with privacy expectations, we propose to operationalize contextual integrity (CI), a framework that equates privacy with the appropriate flow of information in a given context. In particular, we design and evaluate a number of strategies to steer assistants' information-sharing actions to be CI compliant. Our evaluation is based on a novel form filling benchmark composed of human annotations of common webform applications, and it reveals that prompting frontier LLMs to perform CI-based reasoning yields strong results.

The Bicameral Cache is a cache organization proposal for a vector architecture that segregates data according to their access type, distinguishing scalar from vector references. Its aim is to avoid both types of references from interfering in each other's data locality, with a special focus on prioritizing the performance on vector references. The proposed system incorporates an additional, non-polluting prefetching mechanism to help populate the long vector cache lines in advance to increase the hit rate by further exploiting the spatial locality on vector data. Its evaluation was conducted on the Cavatools simulator, comparing the performance to a standard conventional cache, over different typical vector benchmarks for several vector lengths. The results proved the proposed cache speeds up performance on stride-1 vector benchmarks, while hardly impacting non-stride-1's. In addition, the prefetching feature consistently provided an additional value., Comment: 9 pages, 5 figures

Generative Artificial Intelligence has grown exponentially as a result of Large Language Models (LLMs). This has been possible because of the impressive performance of deep learning methods created within the field of Natural Language Processing (NLP) and its subfield Natural Language Generation (NLG), which is the focus of this paper. Within the growing LLM family are the popular GPT-4, Bard and more specifically, tools such as ChatGPT have become a benchmark for other LLMs when solving most of the tasks involved in NLG research. This scenario poses new questions about the next steps for NLG and how the field can adapt and evolve to deal with new challenges in the era of LLMs. To address this, the present paper conducts a review of a representative sample of surveys recently published in NLG. By doing so, we aim to provide the scientific community with a research roadmap to identify which NLG aspects are still not suitably addressed by LLMs, as well as suggest future lines of research that should be addressed going forward.

Pediatric central nervous system tumors are the leading cause of cancer-related deaths in children. The five-year survival rate for high-grade glioma in children is less than 20%. The development of new treatments is dependent upon multi-institutional collaborative clinical trials requiring reproducible and accurate centralized response assessment. We present the results of the BraTS-PEDs 2023 challenge, the first Brain Tumor Segmentation (BraTS) challenge focused on pediatric brain tumors. This challenge utilized data acquired from multiple international consortia dedicated to pediatric neuro-oncology and clinical trials. BraTS-PEDs 2023 aimed to evaluate volumetric segmentation algorithms for pediatric brain gliomas from magnetic resonance imaging using standardized quantitative performance evaluation metrics employed across the BraTS 2023 challenges. The top-performing AI approaches for pediatric tumor analysis included ensembles of nnU-Net and Swin UNETR, Auto3DSeg, or nnU-Net with a self-supervised framework. The BraTSPEDs 2023 challenge fostered collaboration between clinicians (neuro-oncologists, neuroradiologists) and AI/imaging scientists, promoting faster data sharing and the development of automated volumetric analysis techniques. These advancements could significantly benefit clinical trials and improve the care of children with brain tumors.

This paper addresses the regulation and trajectory-tracking problems for two classes of weakly coupled electromechanical systems. To this end, we formulate an energy-based model for these systems within the port-Hamiltonian framework. Then, we employ Lyapunov theory and the notion of contractive systems to develop control approaches in the port-Hamiltonian framework. Remarkably, these control methods eliminate the need for solving partial differential equations or implementing any change of coordinates and are endowed with a physical interpretation. We also investigate the effect of coupled damping on the transient performance and convergence rate of the closed-loop system. Finally, the applicability of the proposed approaches is illustrated in two applications of electromechanical systems via simulations.

Hyperspherical Prototypical Learning (HPL) is a supervised approach to representation learning that designs class prototypes on the unit hypersphere. The prototypes bias the representations to class separation in a scale invariant and known geometry. Previous approaches to HPL have either of the following shortcomings: (i) they follow an unprincipled optimisation procedure; or (ii) they are theoretically sound, but are constrained to only one possible latent dimension. In this paper, we address both shortcomings. To address (i), we present a principled optimisation procedure whose solution we show is optimal. To address (ii), we construct well-separated prototypes in a wide range of dimensions using linear block codes. Additionally, we give a full characterisation of the optimal prototype placement in terms of achievable and converse bounds, showing that our proposed methods are near-optimal., Comment: 14 pages: 9 of the main paper, 2 of references, and 3 of appendices. To appear in the Proceedings of the Geometry-grounded Representation Learning and Generative Modeling at the 41st International Conference on Machine Learning, Vienna, Austria. Code is available at: https://github.com/martinlindstrom/coding_theoretic_hpl

In this article, we investigate a growing trend in the worldwide Quantum Technology (QT) education landscape, that of the development of masters programs, intended to provide graduates with the knowledge and skills to take a job in the quantum industry, while serving a much shorter timeline than a doctoral degree. Through a global survey, we identified 86 masters programs, with substantial growth since 2021. Over time masters have become increasingly interdisciplinary, organised by multiple faculties or through joint degree programs, and offer more hands-on experiences such as internships in companies. Information from program organisers and websites suggests that the intended career destinations of their graduates are a diverse range of industries, and therefore masters programs may be a boon to the industrialisation of quantum technologies. Finally, we identify a range of national efforts to grow the quantum workforce of many countries, quantum program enhancements, which augment the content of existing study programs with quantum content. This may further contribute to the growth and viability of masters programs as a route to the quantum industry.

High kinetic inductance superconductors are gaining increasing interest for the realisation of qubits, amplifiers and detectors. Moreover, thanks to their high impedance, quantum buses made of such materials enable large zero-point fluctuations of the voltage, boosting the coupling rates to spin and charge qubits. However, fully exploiting the potential of disordered or granular superconductors is challenging, as their inductance and, therefore, impedance at high values are difficult to control. Here we have integrated a granular aluminium resonator, having a characteristic impedance exceeding the resistance quantum, with a germanium double quantum dot and demonstrate strong charge-photon coupling with a rate of $g_\text{c}/2\pi= (566 \pm 2)$ MHz. This was achieved due to the realisation of a wireless ohmmeter, which allows \emph{in situ} measurements during film deposition and, therefore, control of the kinetic inductance of granular aluminium films. Reproducible fabrication of circuits with impedances (inductances) exceeding 13 k$\Omega$ (1 nH per square) is now possible. This broadly applicable method opens the path for novel qubits and high-fidelity, long-distance two-qubit gates.

While synthetic tabular data generation using Deep Generative Models (DGMs) offers a compelling solution to data scarcity and privacy concerns, their effectiveness relies on substantial training data, often unavailable in real-world applications. This paper addresses this challenge by proposing a novel methodology for generating realistic and reliable synthetic tabular data with DGMs in limited real-data environments. Our approach proposes several ways to generate an artificial inductive bias in a DGM through transfer learning and meta-learning techniques. We explore and compare four different methods within this framework, demonstrating that transfer learning strategies like pre-training and model averaging outperform meta-learning approaches, like Model-Agnostic Meta-Learning, and Domain Randomized Search. We validate our approach using two state-of-the-art DGMs, namely, a Variational Autoencoder and a Generative Adversarial Network, to show that our artificial inductive bias fuels superior synthetic data quality, as measured by Jensen-Shannon divergence, achieving relative gains of up to 50\% when using our proposed approach. This methodology has broad applicability in various DGMs and machine learning tasks, particularly in areas like healthcare and finance, where data scarcity is often a critical issue., Comment: 19 pages, 6 Figures