Search

We investigate the effectiveness of Explainable AI (XAI) in verifying Machine Unlearning (MU) within the context of harbor front monitoring, focusing on data privacy and regulatory compliance. With the increasing need to adhere to privacy legislation such as the General Data Protection Regulation (GDPR), traditional methods of retraining ML models for data deletions prove impractical due to their complexity and resource demands. MU offers a solution by enabling models to selectively forget specific learned patterns without full retraining. We explore various removal techniques, including data relabeling, and model perturbation. Then, we leverage attribution-based XAI to discuss the effects of unlearning on model performance. Our proof-of-concept introduces feature importance as an innovative verification step for MU, expanding beyond traditional metrics and demonstrating techniques' ability to reduce reliance on undesired patterns. Additionally, we propose two novel XAI-based metrics, Heatmap Coverage (HC) and Attention Shift (AS), to evaluate the effectiveness of these methods. This approach not only highlights how XAI can complement MU by providing effective verification, but also sets the stage for future research to enhance their joint integration., Comment: ICPRW2024

Handling heterogeneity and unpredictability are two core problems in pervasive computing. The challenge is to seamlessly integrate devices with varying computational resources in a dynamic environment to form a cohesive system that can fulfill the needs of all participants. Existing work on systems that adapt to changing requirements typically focuses on optimizing individual variables or low-level Service Level Objectives (SLOs), such as constraining the usage of specific resources. While low-level control mechanisms permit fine-grained control over a system, they introduce considerable complexity, particularly in dynamic environments. To this end, we propose drawing from Active Inference (AIF), a neuroscientific framework for designing adaptive agents. Specifically, we introduce a conceptual agent for heterogeneous pervasive systems that permits setting global systems constraints as high-level SLOs. Instead of manually setting low-level SLOs, the system finds an equilibrium that can adapt to environmental changes. We demonstrate the viability of AIF agents with an extensive experiment design, using heterogeneous and lifelong federated learning as an application scenario. We conduct our experiments on a physical testbed of devices with different resource types and vendor specifications. The results provide convincing evidence that an AIF agent can adapt a system to environmental changes. In particular, the AIF agent can balance competing SLOs in resource heterogeneous environments to ensure up to 98% fulfillment rate., Comment: 11 pages, double column, 15 figures, 2 tables

We develop formal privacy mechanisms for releasing statistics from data with many outlying values, such as income data. These mechanisms ensure that a per-record differential privacy guarantee degrades slowly in the protected records' influence on the statistics being released. Formal privacy mechanisms generally add randomness, or "noise," to published statistics. If a noisy statistic's distribution changes little with the addition or deletion of a single record in the underlying dataset, an attacker looking at this statistic will find it plausible that any particular record was present or absent, preserving the records' privacy. More influential records -- those whose addition or deletion would change the statistics' distribution more -- typically suffer greater privacy loss. The per-record differential privacy framework quantifies these record-specific privacy guarantees, but existing mechanisms let these guarantees degrade rapidly (linearly or quadratically) with influence. While this may be acceptable in cases with some moderately influential records, it results in unacceptably high privacy losses when records' influence varies widely, as is common in economic data. We develop mechanisms with privacy guarantees that instead degrade as slowly as logarithmically with influence. These mechanisms allow for the accurate, unbiased release of statistics, while providing meaningful protection for highly influential records. As an example, we consider the private release of sums of unbounded establishment data such as payroll, where our mechanisms extend meaningful privacy protection even to very large establishments. We evaluate these mechanisms empirically and demonstrate their utility.

The current scenario of IoT is witnessing a constant increase on the volume of data, which is generated in constant stream, calling for novel architectural and logical solutions for processing it. Moving the data handling towards the edge of the computing spectrum guarantees better distribution of load and, in principle, lower latency and better privacy. However, managing such a structure is complex, especially when requirements, also referred to Service Level Objectives (SLOs), specified by applications' owners and infrastructure managers need to be ensured. Despite the rich number of proposals of Machine Learning (ML) based management solutions, researchers and practitioners yet struggle to guarantee long-term prediction and control, and accurate troubleshooting. Therefore, we present a novel ML paradigm based on Active Inference (AIF) -- a concept from neuroscience that describes how the brain constantly predicts and evaluates sensory information to decrease long-term surprise. We implement it and evaluate it in a heterogeneous real stream processing use case, where an AIF-based agent continuously optimizes the fulfillment of three SLOs for three autonomous driving services running on multiple devices. The agent used causal knowledge to gradually develop an understanding of how its actions are related to requirements fulfillment, and which configurations to favor. Through this approach, our agent requires up to thirty iterations to converge to the optimal solution, showing the capability of offering accurate results in a short amount of time. Furthermore, thanks to AIF and its causal structures, our method guarantees full transparency on the decision making, making the interpretation of the results and the troubleshooting effortless.

In space applications, space charge inside dielectric materials can lead to unwanted phenomena such as breakdown and electrostatic discharge. To prevent such incidents, the pulsed electroacoustic (PEA) device is employed to measure the space charge of dielectric samples. Unfortunately, most of the current PEA signal processing methods are inaccurate in studying the dielectric materials that attenuate and disperse the acoustic wave, as well as studying multi-layer or thin samples. A model under the PSpice software is developed in this work with the aim of improving the PEA signal processing. This model is an evolution of previous works, including the attenuation and dispersion of a polymer sample using the nearly local Kramers--Kronig relation. This model of the polymer sample is implemented into PSpice as a frequency dependent transmission line using Branin's method of characteristics model. A comparison of a model simulation and an experimental PEA signal obtained from a PTFE sample is presented.

In this paper we analyze and compare different movement sensors: micro-chip gesture-ID, leap motion, noitom mocap, and specially developed sensor for tapping and foot motion analysis. The main goal is to evaluate the accu-racy of measurements provided by the sensors. This study presents rele-vance, for instance, in tremor/Parkinson disease analysis as well as no touch mechanisms for activation and control of devices. This scenario is especially interesting in COVID-19 scenario. Removing the need to touch a surface, the risk of contagion is reduced., Comment: In: Esposito, A., Faundez-Zanuy, M., Morabito, F.C., Pasero, E. (eds) Applications of Artificial Intelligence and Neural Systems to Data Science. Smart Innovation, Systems and Technologies, vol 360. Springer, Singapore

Ocean mesoscale variability, a key component of the climate system, influences ocean circulation and heat, gas, carbon and nutrient distribution. Trends on Eddy Kinetic Energy (EKE), a metric measuring its intensity, are investigated using two products constructed from 30 years of altimetric observations. Statistically significant positive trends in globally-averaged EKE reveal a strengthening of 1-3% per decade. Regions of intense mesoscale activity become more energetic than other areas. Robust positive EKE trends are observed in the Kuroshio Extension and the Gulf Stream, with remarkable EKE increases of ~50% and ~20%, respectively, over the last decade. Our study opens a new question into how the observed Gulf Stream strengthening impacts the AMOC, and challenges existing climate models emphasizing the necessity for improved small-scale ocean process representation.

High-quality motion reconstruction that follows the user's movements can be achieved by high-end mocap systems with many sensors. However, obtaining such animation quality with fewer input devices is gaining popularity as it brings mocap closer to the general public. The main challenges include the loss of end-effector accuracy in learning-based approaches, or the lack of naturalness and smoothness in IK-based solutions. In addition, such systems are often finely tuned to a specific number of trackers and are highly sensitive to missing data e.g., in scenarios where a sensor is occluded or malfunctions. In response to these challenges, we introduce DragPoser, a novel deep-learning-based motion reconstruction system that accurately represents hard and dynamic on-the-fly constraints, attaining real-time high end-effectors position accuracy. This is achieved through a pose optimization process within a structured latent space. Our system requires only one-time training on a large human motion dataset, and then constraints can be dynamically defined as losses, while the pose is iteratively refined by computing the gradients of these losses within the latent space. To further enhance our approach, we incorporate a Temporal Predictor network, which employs a Transformer architecture to directly encode temporality within the latent space. This network ensures the pose optimization is confined to the manifold of valid poses and also leverages past pose data to predict temporally coherent poses. Results demonstrate that DragPoser surpasses both IK-based and the latest data-driven methods in achieving precise end-effector positioning, while it produces natural poses and temporally coherent motion. In addition, our system showcases robustness against on-the-fly constraint modifications, and exhibits exceptional adaptability to various input configurations and changes.

To the best of our knowledge, a complete characterization of the domains that escape the famous Arrow's impossibility theorem remains an open question. We believe that different ways of proving Arrovian theorems illuminate this problem. This paper presents a new combinatorial topology proof of Arrow's theorem. In PODC 2022, Rajsbaum and Ravent\'os-Pujol proved this theorem using a combinatorial topology approach. This approach uses simplicial complexes to represent the sets of profiles of preferences and that of single preferences. These complexes grow in dimension with the number of alternatives. This makes it difficult to think about the geometry of Arrow's theorem when there are (any) finite number of voters and alternatives. Rajsbaum and Ravent\'os-Pujol (2022) use their combinatorial topology approach only for the base case of two voters and three alternatives and then proceed by induction to prove the general version. The problem with this strategy is that it is unclear how to study domain restrictions in the general case by focusing on the base case and then using induction. Instead, the present article uses the two-dimensional structure of the high-dimensional simplicial complexes (formally, the $2$$\unicode{x2013}$skeleton), yielding a new combinatorial topology proof of this theorem. Moreover, we do not assume the unrestricted domain, but a domain restriction that we call the class of polarization and diversity over triples, which includes the unrestricted domain. By doing so, we obtain a new generalization of Arrow's theorem. This shows that the combinatorial topology approach can be used to study domain restrictions in high dimensions through the $2$$\unicode{x2013}$skeleton.

Over the past six decades, the computing systems field has experienced significant transformations, profoundly impacting society with transformational developments, such as the Internet and the commodification of computing. Underpinned by technological advancements, computer systems, far from being static, have been continuously evolving and adapting to cover multifaceted societal niches. This has led to new paradigms such as cloud, fog, edge computing, and the Internet of Things (IoT), which offer fresh economic and creative opportunities. Nevertheless, this rapid change poses complex research challenges, especially in maximizing potential and enhancing functionality. As such, to maintain an economical level of performance that meets ever-tighter requirements, one must understand the drivers of new model emergence and expansion, and how contemporary challenges differ from past ones. To that end, this article investigates and assesses the factors influencing the evolution of computing systems, covering established systems and architectures as well as newer developments, such as serverless computing, quantum computing, and on-device AI on edge devices. Trends emerge when one traces technological trajectory, which includes the rapid obsolescence of frameworks due to business and technical constraints, a move towards specialized systems and models, and varying approaches to centralized and decentralized control. This comprehensive review of modern computing systems looks ahead to the future of research in the field, highlighting key challenges and emerging trends, and underscoring their importance in cost-effectively driving technological progress., Comment: Preprint submitted to Telematics and Informatics Reports, Elsevier (2024)

Temporal Action Localization (TAL) is a complex task that poses relevant challenges, particularly when attempting to generalize on new -- unseen -- domains in real-world applications. These scenarios, despite realistic, are often neglected in the literature, exposing these solutions to important performance degradation. In this work, we tackle this issue by introducing, for the first time, an approach for Unsupervised Domain Adaptation (UDA) in sparse TAL, which we refer to as Semantic Adversarial unsupervised Domain Adaptation (SADA). Our contributions are threefold: (1) we pioneer the development of a domain adaptation model that operates on realistic sparse action detection benchmarks; (2) we tackle the limitations of global-distribution alignment techniques by introducing a novel adversarial loss that is sensitive to local class distributions, ensuring finer-grained adaptation; and (3) we present a novel set of benchmarks based on EpicKitchens100 and CharadesEgo, that evaluate multiple domain shifts in a comprehensive manner. Our experiments indicate that SADA improves the adaptation across domains when compared to fully supervised state-of-the-art and alternative UDA methods, attaining a performance boost of up to 6.14% mAP.

In the new research and innovation (R&I) paradigm, aimed at a transformation towards more sustainable, inclusive and fair pathways to address societal and environmental challenges, and at generating new patterns of specialisation and new trajectories for socioeconomic development, it is essential to provide monitoring systems and tools to map and understand the contribution of R&I policies and projects. To address this transformation, we present the RIS3-MCAT platform, the result of a line of work aimed at exploring the potential of open data, semantic analysis, and data visualisation, for monitoring challenge-oriented smart specialisation in Catalonia. RIS3-MCAT is an interactive platform that facilitates access to R&I project data in formats that allow for sophisticated analyses of a large volume of texts, enabling the detailed study of thematic specialisations and challenges beyond classical classification systems. Its conceptualisation, development framework and use are presented in this paper., Comment: Accepted and presented at the 27th International Conference on Science, Technology and Innovation Indicators (STI 2023)

FLUKA is among the general-purpose codes for the Monte Carlo simulation of radiation transport that are routinely employed to estimate the production of single-event-upsets (SEUs) in commercial static random access memories (SRAMs) exposed to radiation. Earlier studies concerning the production of SEUs in commercial SRAMs under proton irradiation revealed very good agreement between experimental measurements and FLUKA estimates of the SEU production cross section for proton energies above 20-30 MeV. However, at lower proton energies, where the cross section for SEU production in such low-critical-charge components increases drastically, a FLUKA underestimation of up to two orders of magnitude was observed. Preliminary analyses indicated that this underestimation was in great measure due to the lack of nuclear elastic scattering of protons below 10 MeV in FLUKA up to version 4-3.4. To overcome this limitation, a new model for the nuclear elastic scattering of protons has been developed, combining partial-wave analyses and experimental angular distributions. This newly developed model has been included in FLUKA v4-4.0, and leads to an order-of-magnitude improvement in the agreement between FLUKA and experimental cross sections for the production of SEUs in SRAMs under proton irradiation in the 1-10 MeV energy domain., Comment: 14 pages, 15 figures

Molybdenum disulfide, $MoS_2$, is a next-generation semiconductor and is frequently integrated into emergent optoelectronic technologies based on two-dimensional materials. Here, we present a method that provides direct optical feedback on the thickness and crystallinity of sputter-deposited $MoS_2$ down to the few-layer regime. This smart sensing enables tracking the material's functional properties, such as excitonic response, sheet resistance, and hardness across the amorphous-crystalline transition. To illustrate the potential of such feedback-controlled fabrication, we realized $MoS_2$-based hyperbolic metamaterials (HMM) with controllable optical topological transitions and hardness., Comment: 17 pages, 4 figures, 68 references. Supplementary available with 16 pages, 8 figures, and 10 references

Converging Zero Trust (ZT) with learning techniques can solve various operational and security challenges in Distributed Computing Continuum Systems (DCCS). Implementing centralized ZT architecture is seen as unsuitable for the computing continuum (e.g., computing entities with limited connectivity and visibility, etc.). At the same time, implementing decentralized ZT in the computing continuum requires understanding infrastructure limitations and novel approaches to enhance resource access management decisions. To overcome such challenges, we present a novel learning-driven ZT conceptual architecture designed for DCCS. We aim to enhance ZT architecture service quality by incorporating lightweight learning strategies such as Representation Learning (ReL) and distributing ZT components across the computing continuum. The ReL helps to improve the decision-making process by predicting threats or untrusted requests. Through an illustrative example, we show how the learning process detects and blocks the requests, enhances resource access control, and reduces network and computation overheads. Lastly, we discuss the conceptual architecture, processes, and provide a research agenda.

Machine Learning (ML) is a common tool to interpret and predict the behavior of distributed computing systems, e.g., to optimize the task distribution between devices. As more and more data is created by Internet of Things (IoT) devices, data processing and ML training are carried out by edge devices in close proximity. To ensure Quality of Service (QoS) throughout these operations, systems are supervised and dynamically adapted with the help of ML. However, as long as ML models are not retrained, they fail to capture gradual shifts in the variable distribution, leading to an inaccurate view of the system state. Moreover, as the prediction accuracy decreases, the reporting device should actively resolve uncertainties to improve the model's precision. Such a level of self-determination could be provided by Active Inference (ACI) -- a concept from neuroscience that describes how the brain constantly predicts and evaluates sensory information to decrease long-term surprise. We encompassed these concepts in a single action-perception cycle, which we implemented for distributed agents in a smart manufacturing use case. As a result, we showed how our ACI agent was able to quickly and traceably solve an optimization problem while fulfilling QoS requirements.

Compute Continuum (CC) systems comprise a vast number of devices distributed over computational tiers. Evaluating business requirements, i.e., Service Level Objectives (SLOs), requires collecting data from all those devices; if SLOs are violated, devices must be reconfigured to ensure correct operation. If done centrally, this dramatically increases the number of devices and variables that must be considered, while creating an enormous communication overhead. To address this, we (1) introduce a causality filter based on Markov blankets (MB) that limits the number of variables that each device must track, (2) evaluate SLOs decentralized on a device basis, and (3) infer optimal device configuration for fulfilling SLOs. We evaluated our methodology by analyzing video stream transformations and providing device configurations that ensure the Quality of Service (QoS). The devices thus perceived their environment and acted accordingly -- a form of decentralized intelligence.

In the last few years there is an unprecedented progress in the increase of the power conversion efficiency of perovskite solar cells. Evidently, further advances of the efficiency of these devices will depend on the constraints imposed by the optical and electronic properties of their constituents. Quite apparently that during the manufacturing process of a solar cell, there is an inevitable variation in the thicknesses of various functional layers, which affects the optoelectronic characteristics of the final sample. In this work a possible strategy of the analysis of the solar cell performance is suggested, based on statistically averaging procedure of experimental data. We present a case study, in which the optoelectronic properties of the meso-structured perovskite solar cell (with a mesoporous TiO$_2$ layer) are analysed within the method providing a deeper understanding of the device operation. This method enables an assessment of the overall quality of the device, pointing pathways towards the maximum efficiency design of a perovskite solar cell by material properties tuning., Comment: 13 pages, 10 figures

The physics of skyrmions, and in particular the issue of how to isolate and manipulate them individually, is a subject of major importance nowadays in the community of magnetism. In this article we present an in-depth extension of a study on this issue that was recently proposed by some of the authors [H. D. Rosales, et al. Phys. Rev. Lett. \textbf{130}, 106703 (2023)]. More precisely, we analyse the competition between skyrmions and a chiral spin liquid in a model on the kagome lattice. We first present an analytical overview of the low-energy states using the Luttinger-Tisza approximation. We then study the effect of thermal fluctuations thanks to large-scale Monte-Carlo simulations, and explore the entire parameter space with a magnetic field $B$, in-plane $D^{xy}$ and out-of-plane $D^z$ Dzyaloshinskii-Moriya interactions. While skyrmions and the chiral spin liquid live in different regions of the parameter space, we show how to bring them together, stabilizing a skyrmion fluid in between; a region where the density of well-defined skyrmions can be tuned before obtaining an ordered phase. We investigate in particular the melting of the skyrmion solid. Our analysis also brings to light a long-range ordered phase with Z$_3$ symmetry. At last, we initiate the study of this rich magnetic background on conduction electrons that are coupled to the local spins. We study how the different chiral magnetic textures stabilized in this model (skyrmion solid, liquid and gas and chiral spin liquid) induce a topological Quantum Hall effect. We observe in the ordered skyrmion phase the appearance of Landau levels which persist even in the skyrmion-liquid regime and gradually disappear as the skyrmion density decreases to form a gas., Comment: 19 pages, 16 figures

We consider the problem of the private release of statistics (like aggregate payrolls) where it is critical to preserve the contribution made by a small number of outlying large entities. We propose a privacy formalism, per-record zero concentrated differential privacy (PzCDP), where the privacy loss associated with each record is a public function of that record's value. Unlike other formalisms which provide different privacy losses to different records, PRzCDP's privacy loss depends explicitly on the confidential data. We define our formalism, derive its properties, and propose mechanisms which satisfy PRzCDP that are uniquely suited to publishing skewed or heavy-tailed statistics, where a small number of records contribute substantially to query answers. This targeted relaxation helps overcome the difficulties of applying standard DP to these data products., Comment: 14 pages, 5 figures