Search

Smart-city services are typically developed as closed systems within each city's vertical, communicating and interacting with cloud services while remaining isolated within each provider's domain. With the emergence of 5G private domains and the introduction of new M2M services focusing on autonomous systems, there is a shift from the cloud-based approach to a distributed edge computing paradigm, in a \textit{continuum} orchestration. However, an essential component is missing. Current orchestration tools, designed for cloud-based deployments, lack robust workload isolation, fail to meet timing constraints, and are not tailored to the resource-constrained nature of edge devices. Therefore, new orchestration methods are needed to support MEC environments. The work presented in this paper addresses this gap. Based on the real needs of a smart-city testbed - the Aveiro Living Lab-, we developed a set of orchestration components to facilitate the seamless orchestration of both cloud and edge-based services, encompassing both critical and non-critical services. This work extends the current Kubernetes orchestration platform to include a novel location-specific resource definition, a custom scheduler to accommodate real-time and legacy services, continuous service monitoring to detect sub-optimal states, and a refined load balancing mechanism that prioritizes the fastest response times.

With modern vehicles evolving with more features, services, complex systems, with more sensors, actuators, and processing units, it is essential to think about vehicles not only as means of transportation that may tend towards full autonomy, but also as adaptive objects, that suit themselves to the needs of occupants. Vehicular services can be developed to support these adaptations. However, the increasing complexity of vehicular service development, even with current standardizations and best practices and guidelines, are insufficient to tackle the high complexity of development, with expectations of up to 1 (U.S.) billion lines of code for a fully (level 5) autonomous vehicle. Within this survey, the paradigm of Deterministic Software Defined Vehicles is explored towards increasing the quality and easiness of the development of services for automotive. Towards this, a proposed vision with four pillars is also provided: the deterministic network configurator, the data layer configurator, and the hypervisor configurator and the vehicle abstraction layer, all coordinated by a software orchestrator.

This work investigates Artificial Intelligence (AI) systems that detect respiratory insufficiency (RI) by analyzing speech audios, thus treating speech as a RI biomarker. Previous works collected RI data (P1) from COVID-19 patients during the first phase of the pandemic and trained modern AI models, such as CNNs and Transformers, which achieved $96.5\%$ accuracy, showing the feasibility of RI detection via AI. Here, we collect RI patient data (P2) with several causes besides COVID-19, aiming at extending AI-based RI detection. We also collected control data from hospital patients without RI. We show that the considered models, when trained on P1, do not generalize to P2, indicating that COVID-19 RI has features that may not be found in all RI types., Comment: 5 pages, 2 figures, 1 table. Published in Artificial Intelligence in Medicine (AIME) 2023

This paper is devoted to the study of Hopf braces projections in a monoidal setting. Given a cocommutative Hopf brace ${\mathbb H}$ in a strict symmetric monoidal category ${\sf C}$, we define the braided monoidal category of left Yetter-Drinfeld modules over ${\mathbb H}$. For a Hopf brace ${\mathbb A}$ in this category, we introduce the concept of bosonizable Hopf brace and we prove that its bosonization ${\mathbb A}\blacktriangleright\hspace{-0.15cm}\blacktriangleleft {\mathbb H}$ is a new Hopf brace in ${\sf C}$ that gives rise to a projection of Hopf braces satisfying certain properties. Finally, taking these properties into account, we introduce the notions of v$_{i}$-strong projection over ${\mathbb H}$, $i=1,2,3,4$, and we prove that there is a categorical equivalence between the categories of bosonizable Hopf braces in the category of left Yetter-Drinfeld modules over ${\mathbb H}$ and the category of v$_{4}$-strong projections over ${\mathbb H}$.

Transformer-based language models spread FLOPs uniformly across input sequences. In this work we demonstrate that transformers can instead learn to dynamically allocate FLOPs (or compute) to specific positions in a sequence, optimising the allocation along the sequence for different layers across the model depth. Our method enforces a total compute budget by capping the number of tokens ($k$) that can participate in the self-attention and MLP computations at a given layer. The tokens to be processed are determined by the network using a top-$k$ routing mechanism. Since $k$ is defined a priori, this simple procedure uses a static computation graph with known tensor sizes, unlike other conditional computation techniques. Nevertheless, since the identities of the $k$ tokens are fluid, this method can expend FLOPs non-uniformly across the time and model depth dimensions. Thus, compute expenditure is entirely predictable in sum total, but dynamic and context-sensitive at the token-level. Not only do models trained in this way learn to dynamically allocate compute, they do so efficiently. These models match baseline performance for equivalent FLOPS and wall-clock times to train, but require a fraction of the FLOPs per forward pass, and can be upwards of 50\% faster to step during post-training sampling.

In this paper, we introduce the notions of $\alpha$-quasicomplemented and totally $\alpha$-quasicomplemented subspaces and we established some results under these contexts. We show, for example, that if $X$ is a separable or reflexive Banach space and $Y$ is a closed infinite codimensional subspace of $X$, then $Y$ is totally$\mathit{\ }\alpha$-quasicomplemented if, and only if, $\alpha<\aleph_{0}$ $\left( \text{this is an analogue of the theorem of Murray-Mackey and Lindenstrauss}\right) $. We also show that if $H$ is a Hilbert space and $Y,W\subset H$ are closed subspaces of $H$ such that $W$ is orthogonal to $Y$ and $\operatorname{codim}\left( Y+W\right) =\infty$, then $Y$ has a quasicomplement $Z$ containing $W$ with $\dim Z/W=\infty$. Other results in the different contexts are also included. Such results establish a connection between the theory of quasicomplemented subspaces and $\left( \alpha,\beta\right) $-spaceability.

In this paper, we introduce SaulLM-7B, a large language model (LLM) tailored for the legal domain. With 7 billion parameters, SaulLM-7B is the first LLM designed explicitly for legal text comprehension and generation. Leveraging the Mistral 7B architecture as its foundation, SaulLM-7B is trained on an English legal corpus of over 30 billion tokens. SaulLM-7B exhibits state-of-the-art proficiency in understanding and processing legal documents. Additionally, we present a novel instructional fine-tuning method that leverages legal datasets to further enhance SaulLM-7B's performance in legal tasks. SaulLM-7B is released under the MIT License.

We report the first experimental demonstration of the replica symmetry breaking (RSB) phenomenon in a fiber laser system supporting standard mode-locking (SML) regime. Though theoretically predicted, this photonic glassy phase remained experimentally undisclosed so far. We employ an ytterbium-based mode-locked fiber laser with a very rich phase diagram. Two phase transitions are observed separating three different regimes: cw, quasi-mode-locking (QML), and SML. The regimes are intrinsically related to the distinct dynamics of intensity fluctuations in the laser spectra. We set the connection between the RSB glassy phase with frustrated modes and onset of L-shaped intensity distributions in the QML regime, which impact directly the replica overlap measure., Comment: 17 pages, 12 figures, 1 Supplementary material

Current driven domain wall motion in curved Heavy Metal/Ferrimagnetic/Oxide multilayer strips is investigated using systematic micromagnetic simulations which account for spin-orbit coupling phenomena. Domain wall velocity and characteristic relaxation times are studied as functions of the geometry, curvature and width of the strip, at and out of the angular momentum compensation. Results show that domain walls can propagate faster and without a significant distortion in such strips in contrast to their ferromagnetic counterparts. Using an artificial system based on a straight strip with an equivalent current density distribution, we can discern its influence on the wall terminal velocity, as part of a more general geometrical influence due to the curved shape. Curved and narrow ferrimagnetic strips are promising candidates for designing high speed and fast response spintronic circuitry based on current-driven domain wall motion.

Nucleation of domain walls by current-driving a single domain wall, confined to the junction area of two symmetrical strips, is investigated using systematic micromagnetic simulations. Secondary domain walls (equivalently, bits encoded in domains) are simultaneously nucleated and driven by alternatively applying current pulses between two terminals in the structure. Simulations show that nanosecond-duration current pulses nucleate and drive series of robust up/down domains even under realistic conditions. These results demonstrate a technique for sequentially nucleating and shifting domain walls without using attached external bit lines, fields or modifying the ferromagnetic strip., Comment: 11 pages, 3 figures

Spin glass theory, as a paradigm for describing disordered magnetic systems, constitutes a prominent subject of study within statistical physics. Replica symmetry breaking (RSB), as one of the pivotal concepts for the understanding of spin glass theory, means that, under identical conditions disordered systems can yield distinct states with nontrivial correlations. Random fiber laser (RFL) based on Rayleigh scattering (RS) is a complex disordered system, owing to the disorder and stochasticity of RS. In this work, for the first time, we elaborate a precise theoretical model for studying the photonic phase transition via the platform of RS-based RFL, in which we clearly reveal that, apart from the pump power, the photon phase variation in RFL is also an analogy to the temperature term in spin glass phase transition, leading to a novel insight into the intrinsic mechanisms of photonic phase transition. In addition, based on this model and real-time high-fidelity detection spectral evolution, we theoretically predict and experimentally observe the mode-asymmetric characteristics of photonic phase transition in RS-based RFL. This finding contributes to a deeper understanding of the photonic RSB regime and the dynamics of RS-based RFL., Comment: 15 pages, 9 figures

In this paper we introduce the notion of generalized invertible 1-cocycle in a strict braided monoidal category C, and we prove that the category of Hopf trusses is equivalent to the category of generalized invertible 1-cocycles. On the other hand, we also introduce the notions of module for a Hopf truss and for a generalized invertible 1-cocycle. We prove some functorial results involving these categories of modules and we show that the category of modules associated to a generalized invertible 1-cocycle is equivalent to a category of modules associated to a suitable Hopf truss. Finally, assuming that in C we have equalizers, we introduce the notion of Hopf-module in the Hopf truss setting and we obtain the Fundamental Theorem of Hopf modules associated to a Hopf truss.

The rapid proliferation of smart cities has transformed urban landscapes into dynamic ecosystems teeming with interconnected computational nodes and sensors. During this evolution, the search for seamless communication in time-critical scenarios has become evident. With the escalating complexity of urban environments, envisioning a future with a blend of autonomous and conventional systems, each demanding distinct quality-of-service considerations, services in smart cities vary criticality levels and necessitate differentiated traffic handling, prioritizing critical flows without compromising the network's reliability or failing on hard real-time requirements. To tackle these challenges, in this article we propose a Time-Sensitive Networking (TSN) approach which, at the scale of a smart city network, presents multifaceted challenges, notably interoperability among diverse technologies and standards. Nonetheless, TSN emerges as a promising toolkit, encompassing synchronization, latency management, redundancy, and configuration functionalities crucial for addressing smart city challenges. Moreover, the article scrutinizes how TSN, predominantly utilized in domains like automotive and industry, can be tailored to suit the intricate needs of smart cities, emphasizing the necessity for adaptability and scalability in network design. This survey consolidates current research on TSN, outlining its potential in fortifying critical machine-to-machine communications within smart cities while highlighting future challenges, potential solutions, and a roadmap for integrating TSN effectively into the fabric of urban connectivity.

The two-particle finite-volume scattering formalism derived by L\"uscher and generalized in many subsequent works does not hold for energies far enough below the two-particle threshold to reach the nearest left-hand cut. The breakdown of the formalism is signaled by the fact that a real scattering amplitude is predicted in a regime where it should be complex. In this work, we address this limitation by deriving an extended formalism that includes the nearest branch cut, arising from single particle exchange. We focus on two-nucleon ($NN \to NN$) scattering, for which the cut arises from pion exchange, but give expressions for any system with a single channel of identical particles. The new result takes the form of a modified quantization condition that can be used to constrain an intermediate K-matrix in which the cut is removed. In a second step, integral equations, also derived in this work, must be used to convert the K-matrix to the physical scattering amplitude. We also show how the new formalism reduces to the standard approach when the $N \to N \pi$ coupling is set to zero., Comment: 60 pages, 9 figures, v2: fix typos and improve discussion of integral equations, additional figure to better explain time-ordered perturbation theory

In this note we generalize a criterion within the concept of infinite dense lineability due to Calder\'on-Moreno, Gerlach-Mena and Prado-Bassas. We also introduce and explore some local notions of lineability.

In this paper we introduce in a braided setting the notion of left module for an invertible 1-cocycle and we prove some categorical equivalences between categories of modules associated to an invertible 1-cocycle and categories of modules associated to Hopf braces.