Search

Electroencephalogram (EEG) has emerged as a cost-effective and efficient method for supporting neurologists in assessing Alzheimer's disease (AD). Existing approaches predominantly utilize handcrafted features or Convolutional Neural Network (CNN)-based methods. However, the potential of the transformer architecture, which has shown promising results in various time series analysis tasks, remains underexplored in interpreting EEG for AD assessment. Furthermore, most studies are evaluated on the subject-dependent setup but often overlook the significance of the subject-independent setup. To address these gaps, we present ADformer, a novel multi-granularity transformer designed to capture temporal and spatial features to learn effective EEG representations. We employ multi-granularity data embedding across both dimensions and utilize self-attention to learn local features within each granularity and global features among different granularities. We conduct experiments across 5 datasets with a total of 525 subjects in setups including subject-dependent, subject-independent, and leave-subjects-out. Our results show that ADformer outperforms existing methods in most evaluations, achieving F1 scores of 75.19% and 93.58% on two large datasets with 65 subjects and 126 subjects, respectively, in distinguishing AD and healthy control (HC) subjects under the challenging subject-independent setup., Comment: 17 pages main paper + 3 pages supplementary materials. This work will submit to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

Coupling between relativistic jets launched by accreting supermassive black holes and the surrounding gaseous media is a vital ingredient in galaxy evolution models. To constrain the environments in which this feedback takes place over cosmic time, we study the host halo properties of luminous low-frequency radio galaxies ($L_{150 \ \mathrm{MHz}} \gtrsim$ 25.25 W/Hz) selected with the International LOFAR Telescope out to $z \sim 2$ through tomographic clustering and cosmic microwave background lensing measurements. We find that these systems occupy halos characteristic of galaxy groups ($M_h = 10^{13} - 10^{14} h^{-1} M_{\odot}$), evolving at a rate consistent with the mean growth rate of halos over the past $\sim$10 Gyr. The coevolution of the clustering and the luminosity function reveals that the duty cycle of these systems is of order $\sim 10\%$ but has been mildly increasing since $z\sim 2$, while the duty cycle of quasars has been declining. We estimate the characteristic kinetic heating power injected by powerful jets per halo as a function of mass, and compare to the same quantity injected by quasar winds. We find that powerful jet heating dominates over quasar winds in halos $M_h \gtrsim 10^{13} h^{-1} M_{\odot}$ at $z < 2$. These results conform to the paradigm of galaxy evolution in which mechanical jet power feedback is the dominant heating mechanism of the gas content of groups and clusters., Comment: Accepted to ApJ

Digital Twins (DTs) are set to become a key enabling technology in future wireless networks, with their use in network management increasing significantly. We developed a DT framework that leverages the heterogeneity of network access technologies as a resource for enhanced network performance and management, enabling smart data handling in the physical network. Tested in a Campus Area Network environment, our framework integrates diverse data sources to provide real-time, holistic insights into network performance and environmental sensing. We also envision that traditional analytics will evolve to rely on emerging AI models, such as Generative AI (GenAI), while leveraging current analytics capabilities. This capacity can simplify analytics processes through advanced ML models, enabling descriptive, diagnostic, predictive, and prescriptive analytics in a unified fashion. Finally, we present specific research opportunities concerning interoperability aspects and envision aligning advancements in DT technology with evolved AI integration., Comment: A shortened version of this paper is currently under review for publication in an IEEE magazine. If accepted, the copyright will be transferred to IEEE

We present the deepest wide-field 115-166 MHz image at sub-arcsecond resolution spanning an area of 2.5 by 2.5 degrees centred at the ELAIS-N1 deep field. To achieve this, we improved the calibration for the International LOFAR Telescope. This enhancement enabled us to efficiently process 32 hrs of data from four different 8-hr observations using the high-band antennas (HBAs) of all 52 stations, covering baselines up to approximately 2,000 km across Europe. The DI calibration was improved by using an accurate sky model and refining the series of calibration steps on the in-field calibrator, while the DD calibration was improved by adopting a more automated approach for selecting the DD calibrators and inspecting the self-calibration on these sources. We also added an additional round of self-calibration for the Dutch core and remote stations in order to refine the solutions for shorter baselines. To complement our highest resolution at 0.3", we also made intermediate resolution wide-field images at 0.6" and 1.2". Our resulting wide-field images achieve a central noise level of 14 muJy/beam at 0.3", doubling the depth and uncovering four times more objects than the Lockman Hole deep field image at comparable resolution but with only 8 hrs of data. Compared to LOFAR imaging without the international stations, we note that due to the increased collecting area and the absence of confusion noise, we reached a point-source sensitivity comparable to a 500-hr ELAIS-N1 6" image with 16 times less observing time. Importantly, we have found that the computing costs for the same amount of data are almost halved (to about 139,000 CPU hrs per 8 hrs of data) compared to previous efforts, though they remain high. Our work underscores the value and feasibility of exploiting all Dutch and international LOFAR stations to make deep wide-field images at sub-arcsecond resolution., Comment: Images and catalogues will be available on https://lofar-surveys.org/hd-en1.html

On-device inference holds great potential for increased energy efficiency, responsiveness, and privacy in edge ML systems. However, due to less capable ML models that can be embedded in resource-limited devices, use cases are limited to simple inference tasks such as visual keyword spotting, gesture recognition, and predictive analytics. In this context, the Hierarchical Inference (HI) system has emerged as a promising solution that augments the capabilities of the local ML by offloading selected samples to an edge server or cloud for remote ML inference. Existing works demonstrate through simulation that HI improves accuracy. However, they do not account for the latency and energy consumption on the device, nor do they consider three key heterogeneous dimensions that characterize ML systems: hardware, network connectivity, and models. In contrast, this paper systematically compares the performance of HI with on-device inference based on measurements of accuracy, latency, and energy for running embedded ML models on five devices with different capabilities and three image classification datasets. For a given accuracy requirement, the HI systems we designed achieved up to 73% lower latency and up to 77% lower device energy consumption than an on-device inference system. The key to building an efficient HI system is the availability of small-size, reasonably accurate on-device models whose outputs can be effectively differentiated for samples that require remote inference. Despite the performance gains, HI requires on-device inference for all samples, which adds a fixed overhead to its latency and energy consumption. Therefore, we design a hybrid system, Early Exit with HI (EE-HI), and demonstrate that compared to HI, EE-HI reduces the latency by up to 59.7% and lowers the device's energy consumption by up to 60.4%.

As the use of Large Language Models (LLMs) becomes more widespread, understanding their self-evaluation of confidence in generated responses becomes increasingly important as it is integral to the reliability of the output of these models. We introduce the concept of Confidence-Probability Alignment, that connects an LLM's internal confidence, quantified by token probabilities, to the confidence conveyed in the model's response when explicitly asked about its certainty. Using various datasets and prompting techniques that encourage model introspection, we probe the alignment between models' internal and expressed confidence. These techniques encompass using structured evaluation scales to rate confidence, including answer options when prompting, and eliciting the model's confidence level for outputs it does not recognize as its own. Notably, among the models analyzed, OpenAI's GPT-4 showed the strongest confidence-probability alignment, with an average Spearman's $\hat{\rho}$ of 0.42, across a wide range of tasks. Our work contributes to the ongoing efforts to facilitate risk assessment in the application of LLMs and to further our understanding of model trustworthiness., Comment: 9 pages (excluding references), accepted to ACL 2024 Main Conference

Spectral energy distribution (SED) fitting has been extensively used to determine the nature of the faint radio source population. Recent efforts have combined fits from multiple SED-fitting codes to account for the host galaxy and any active nucleus that may be present. We show that it is possible to produce similar-quality classifications using a single energy-balance SED fitting code, Prospector, to model up to 26 bands of UV$-$far-infrared aperture-matched photometry for $\sim$31,000 sources in the ELAIS-N1 field from the LOFAR Two-Metre Sky Survey (LoTSS) Deep fields first data release. One of a new generation of SED-fitting codes, Prospector accounts for potential contributions from radiative active galactic nuclei (AGN) when estimating galaxy properties, including star formation rates (SFRs) derived using non-parametric star formation histories. Combining this information with radio luminosities, we classify 92 per cent of the radio sources as a star-forming galaxy, high/low-excitation radio galaxy, or radio-quiet AGN and study the population demographics as a function of 150 MHz flux density, luminosity, SFR, stellar mass, redshift and apparent $r$-band magnitude. Finally, we use Prospector SED fits to investigate the SFR$-$150 MHz luminosity relation for a sample of $\sim$$133,000~3.6~\mu$m-selected $z<1$ sources, finding that the stellar mass dependence is significantly weaker than previously reported, and may disappear altogether at $\log_{10} (\mathrm{SFR}/M_\odot~\mathrm{yr}^{-1}) > 0.5$. This approach makes it significantly easier to classify radio sources from LoTSS and elsewhere, and may have important implications for future studies of star-forming galaxies at radio wavelengths., Comment: 20 pages, 13 figures, accepted for publication in MNRAS

This article introduces Follow-Me AI, a concept designed to enhance user interactions with smart environments, optimize energy use, and provide better control over data captured by these environments. Through AI agents that accompany users, Follow-Me AI negotiates data management based on user consent, aligns environmental controls as well as user communication and computes resources available in the environment with user preferences, and predicts user behavior to proactively adjust the smart environment. The manuscript illustrates this concept with a detailed example of Follow-Me AI in a smart campus setting, detailing the interactions with the building's management system for optimal comfort and efficiency. Finally, this article looks into the challenges and opportunities related to Follow-Me AI.

The Hierarchical Inference (HI) paradigm employs a tiered processing: the inference from simple data samples are accepted at the end device, while complex data samples are offloaded to the central servers. HI has recently emerged as an effective method for balancing inference accuracy, data processing, transmission throughput, and offloading cost. This approach proves particularly efficient in scenarios involving resource-constrained edge devices, such as IoT sensors and micro controller units (MCUs), tasked with executing tinyML inference. Notably, it outperforms strategies such as local inference execution, inference offloading to edge servers or cloud facilities, and split inference (i.e., inference execution distributed between two endpoints). Building upon the HI paradigm, this work explores different techniques aimed at further optimizing inference task execution. We propose and discuss three distinct HI approaches and evaluate their utility for image classification.

Given a pre-monotone Lagrangian link, we obtain Hofer energy estimates for Hamiltonian diffeomorphisms preserving it. Such estimates depend on the braid type of the Hamiltonian diffeomorphism only, and the natural language to talk about this phenomenon is provided by a family of norms on braid groups for surfaces with boundary. This generalises the results obtained by the first author to higher genus surfaces with boundary., Comment: Exposition has been improved and typos have been corrected. No changes to the main result have been made

We present a novel scheme for rapid quantitative analysis of debris generated during experiments with solid targets following relativistic laser-plasma interaction at high-power laser facilities. Experimental data indicates that predictions by available modelling for non-mass-limited targets are reasonable, with debris on the order of hundreds ug-per-shot. We detect for the first time that several % of the debris is ejected directional following the target normal (rear- and interaction side); and confirm previous work that found the debris ejection in direction of the interaction side to be larger than on the side of the target rear.

Studies show that both radio jets from the active galactic nuclei (AGN) and the star formation (SF) activity in quasar host galaxies contribute to the quasar radio emission; yet their relative contributions across the population remain unclear. Here, we present an improved parametric model that allows us to statistically separate the SF and AGN components in observed quasar radio flux density distributions, and investigate how their relative contributions evolve with AGN bolometric luminosity ($L_\mathrm{bol}$) and redshift ($z$) using a fully Bayesian method. Based on the newest data from LOFAR Two-Metre Sky Survey Data Release 2, our model gives robust fitting results out to $z\sim4$, showing a quasar host galaxy SFR evolution that increases with bolometric luminosity and with redshift out to $z\sim4$. This differs from the global cosmic SFR density, perhaps due to the importance of galaxy mergers. The prevalence of radio AGN emissions increases with quasar luminosity, but has little dependence on redshift. Furthermore, our new methodology and large sample size allow us to subdivide our dataset to investigate the role of other parameters. Specifically, in this paper, we explore quasar colour and demonstrate that the radio excess in red quasars is due to an enhancement in AGN-related emission, since the host galaxy SF contribution to the total radio emission is independent of quasar colour. We also find evidence that this radio enhancement occurs mostly in quasars with weak or intermediate radio power., Comment: Accepted for publication in MNRAS. 19 pages, 13 figures, 1 table

Radio-mode feedback associated with the active galactic nuclei (AGN) at the cores of galaxy clusters injects large amount of energy into the intracluster medium (ICM), offsetting radiative losses through X-ray emission. This mechanism prevents the ICM from rapidly cooling down and fueling extreme starburst activity as it accretes onto the central galaxies, and is therefore a key ingredient in the evolution of galaxy clusters. However, the influence and mode of feedback at high redshifts (z~1) remains largely unknown. Low-frequency sub-arcsecond resolution radio observations taken with the International LOFAR Telescope have demonstrated their ability to assist X-ray observations with constraining the energy output from the AGNs (or "cavity power") in galaxy clusters, thereby enabling research at higher redshifts than before. In this pilot project, we test this hybrid method on a high redshift (0.6

The origin of radio emission in different populations of radio-quiet quasars is relatively unknown, but recent work has uncovered various drivers of increased radio-detection fraction. In this work, we pull together three known factors: optical colour ($g-i$), \CIV Distance (a proxy for $L/L_{Edd}$) and whether or not the quasar contains broad absorption lines (BALQSOs) which signify an outflow. We use SDSS DR14 spectra along with the LOFAR Two Metre Sky Survey Data Release 2 and find that each of these properties have an independent effect. BALQSOs are marginally more likely to be radio-detected than non-BALQSOs at similar colours and $L/L_{Edd}$, moderate reddening significantly increases the radio-detection fraction and the radio-detection increases with $L/L_{Edd}$ above a threshold for all populations. We test a widely used simple model for radio wind shock emission and calculate energetic efficiencies that would be required to reproduce the observed radio properties. We discuss interpretations of these results concerning radio-quiet quasars more generally. We suggest that radio emission in BALQSOs is connected to a different physical origin than the general quasar population since they show different radio properties independent of colour and \CIV distance., Comment: 14 pages, 8 figures, 2 tables

Optogenetic modulation of adenosine triphosphatase (ATPase) expression represents a novel approach to maximize bioprocess efficiency by leveraging enforced adenosine triphosphate (ATP) turnover. In this study, we experimentally implement a model-based open-loop optimization scheme for optogenetic modulation of the expression of ATPase. Increasing the intracellular concentration of ATPase, and thus the level of ATP turnover, in bioprocesses with product synthesis coupled with ATP generation, can lead to increased product formation and substrate uptake. Previous simulation studies formulated optimal control problems using dynamic constraint-based models to find optimal light inputs in fermentations with optogenetically mediated ATPase expression. However, using these models poses challenges due to resulting bilevel optimizations and complex parameterization. Here, we outline a simplified unsegregated and quasi-unstructured kinetic modeling approach that reduces the number of dynamic states and leads to single-level optimizations. The models can be augmented with Gaussian processes to compensate for model uncertainties. We implement optimal control constrained by knowledge-based and hybrid models for optogenetic ATPase expression in $\textit{Escherichia coli}$ with lactate as the main product. To do so, we genetically engineer $\textit{E. coli}$ to obtain optogenetic expression of ATPase using the CcaS/CcaR system. This represents the first experimental implementation of model-based optimization of ATPase expression in bioprocesses., Comment: 24 pages, 5 figures, journal submission

In today's digital world, Generative Artificial Intelligence (GenAI) such as Large Language Models (LLMs) is becoming increasingly prevalent, extending its reach across diverse applications. This surge in adoption has sparked a significant increase in demand for data-centric GenAI models, highlighting the necessity for robust data communication infrastructures. Central to this need are message brokers, which serve as essential channels for data transfer within various system components. This survey aims to delve into a comprehensive analysis of traditional and modern message brokers, offering a comparative study of prevalent platforms. Our study considers numerous criteria including, but not limited to, open-source availability, integrated monitoring tools, message prioritization mechanisms, capabilities for parallel processing, reliability, distribution and clustering functionalities, authentication processes, data persistence strategies, fault tolerance, and scalability. Furthermore, we explore the intrinsic constraints that the design and operation of each message broker might impose, recognizing that these limitations are crucial in understanding their real-world applicability. Finally, this study examines the enhancement of message broker mechanisms specifically for GenAI contexts, emphasizing the criticality of developing a versatile message broker framework. Such a framework would be poised for quick adaptation, catering to the dynamic and growing demands of GenAI in the foreseeable future. Through this dual-pronged approach, we intend to contribute a foundational compendium that can guide future innovations and infrastructural advancements in the realm of GenAI data communication., Comment: 20 pages, 181 references, 7 figures, 5 tables

We present a comprehensive study of the physical origin of radio emission in optical quasars at redshifts z < 2.5. We focus particularly on the associations between compact radio emission, dust reddening, and outflows identified in our earlier work. Leveraging the deepest low-frequency radio data available to date (LoTSS Deep DR1), we achieve radio detection fractions of up to 94%, demonstrating the virtual ubiquity of radio emission in quasars, and a continuous distribution in radio loudness. Through our analysis of radio properties, combined with spectral energy distribution modeling of multiwavelength photometry, we establish that the primary source of radio emission in quasars is the AGN, rather than star formation. Modeling the dust reddening of the accretion disk emission shows a continuous increase in radio detection in quasars as a function of the reddening parameter E(B-V), suggesting a causal link between radio emission and dust reddening. Confirming previous findings, we observe that the radio excess in red quasars is most pronounced for sources with compact radio morphologies and intermediate radio loudness. We find a significant increase in [Oiii] and Civ outflow velocities for red quasars not seen in our control sample, with particularly powerful [Oiii] winds in those around the radio-quiet/radio-loud threshold. Based on the combined characterisation of radio, reddening, and wind properties in our sample, we favor a model in which the compact radio emission observed in quasars originates in compact radio jets and their interaction with a dusty, circumnuclear environment. Our results align with the theory that jet-induced winds and shocks resulting from this interaction are the origin of the enhanced radio emission in red quasars. Further investigation of this model is crucial for advancing our understanding of quasar feedback mechanisms and their role in galaxy evolution., Comment: 11 pages, 8 Figures, submitted to A&A

We present the first deep (72 hours of observations) radio image of the Euclid Deep Field North (EDFN) obtained with the LOw-Frequency ARray (LOFAR) High Band Antenna (HBA) at 144 MHz. The EDFN is the latest addition to the LOFAR Two-Metre Sky Survey (LoTSS) Deep Fields and these observations represent the first data release for this field. The observations produced a 6" resolution image with a central r.m.s. noise of $32\,\mu$Jy\,beam$^{-1}$. A catalogue of $\sim 23,000$ radio sources above a signal-to-noise ratio (SNR) threshold of 5 is extracted from the inner circular 10 deg$^2$ region. We discuss the data analysis and we provide a detailed description of how we derived the catalogue of radio sources and on the issues related to direction-dependent calibration and their effects on the final products. Finally, we derive the radio source counts at 144 MHz in the EDFN using catalogues of mock radio sources to derive the completeness correction factors. The source counts in the EDFN are consistent with those obtained from the first data release of the other LoTSS Deep Fields (ELAIS-N1, Lockman Hole and Bootes), despite the different method adopted to construct the final catalogue and to assess its completeness., Comment: accepted for publication in A&A

Radio-loud active galactic nuclei (RLAGN) play an important role in the evolution of galaxies through the effects on their environment. The two major morphological classes are core-bright (FRI) and edge-bright (FRII) sources. With the LOw-Frequency ARray (LOFAR) we compare the FRI and FRII evolution down to lower flux densities and with larger samples than before with the aim to examine the cosmic space density evolution for FRIs and FRIIs by analyzing their space density evolution between L_150~10^24.5 W/Hz and L_150~10^28.5 W/Hz and up to z=2.5. We construct radio luminosity functions (RLFs) from FRI and FRII catalogues based on recent data from LOFAR at 150MHz to study the space densities as a function of radio luminosity and redshift. To partly correct for selection biases and completeness, we simulate how sources appear at a range of redshifts. We report a space density enhancement from low to high redshift for FRI and FRII sources brighter than L_150~10^27 W/Hz. This is possibly related to the higher gas availability in the earlier denser universe. The constant FRI/FRII space density ratio evolution as a function of radio luminosity and redshift in our results suggests that the jet-disruption of FRIs might be primarily caused by events occurring on scales within the host galaxy, rather than being driven by changes in the overall large-scale environment. Remaining selection biases in our results also highlight the need to resolve more sources at angular scales below 40 arcsec and therefore strengthens the motivation for the further development and automation of the calibration and imaging pipeline of LOFAR data to produce images at sub-arcsecond resolution.

Radio galaxies dominate the radio sky and are essential to the galaxy evolution puzzle. High-resolution studies of statistical samples of radio galaxies are expected to shed light on the triggering mechanisms of the AGN, alternating between the phases of activity and quiescence. In this work, we focus on the sub-arcsec radio structures in the central regions of the 35 radio galaxies over 6.6 $deg^2$ of the Lockman Hole region. These sources were previously classified as active, remnant, and candidate restarted radio galaxies using 150 MHz LOFAR observations. We examine the morphologies and study the spectral properties of their central regions to explore their evolutionary stages and revise the criteria used to select the initial sample. We use the newly available LOFAR 150 MHz image obtained using international baselines, achieving 0.38'' x 0.30'' resolution, making this the first systematic study of the nuclear regions at high resolution and low frequency. We use publicly available images from the FIRST survey at 1.4 GHz and the Karl G. Jansky VLA Sky Survey at 3 GHz to achieve our goals. In addition, for one restarted candidate we present new dedicated observations with the VLA at 3 GHz. We have found various morphologies of the central regions of the radio galaxies in our sample, some resembling miniature double-double radio galaxies. We also see the beginnings of active jets or distinct detections unrelated to the large-scale structure. Furthermore, we have found diverse radio spectra in our sample - flat, steep, or peaked between 150 MHz and 3 GHz, indicative of the different life-cycle phases. Based on these analyses, we confirm five of six previously considered restarted candidates and identify three more from the active sample, supporting previous results suggesting that the restarted phase can occur after a relatively short remnant phase (i.e. a few tens of millions of years)., Comment: Accepted for publication in A&A, 22 pages, 5 tables, 8 figures

The evolution towards 6G architecture promises a transformative shift in communication networks, with artificial intelligence (AI) playing a pivotal role. This paper delves deep into the seamless integration of Large Language Models (LLMs) and Generalized Pretrained Transformers (GPT) within 6G systems. Their ability to grasp intent, strategize, and execute intricate commands will be pivotal in redefining network functionalities and interactions. Central to this is the AI Interconnect framework, intricately woven to facilitate AI-centric operations within the network. Building on the continuously evolving current state-of-the-art, we present a new architectural perspective for the upcoming generation of mobile networks. Here, LLMs and GPTs will collaboratively take center stage alongside traditional pre-generative AI and machine learning (ML) algorithms. This union promises a novel confluence of the old and new, melding tried-and-tested methods with transformative AI technologies. Along with providing a conceptual overview of this evolution, we delve into the nuances of practical applications arising from such an integration. Through this paper, we envisage a symbiotic integration where AI becomes the cornerstone of the next-generation communication paradigm, offering insights into the structural and functional facets of an AI-native 6G network.

The conventional federated learning (FedL) architecture distributes machine learning (ML) across worker devices by having them train local models that are periodically aggregated by a server. FedL ignores two important characteristics of contemporary wireless networks, however: (i) the network may contain heterogeneous communication/computation resources, and (ii) there may be significant overlaps in devices' local data distributions. In this work, we develop a novel optimization methodology that jointly accounts for these factors via intelligent device sampling complemented by device-to-device (D2D) offloading. Our optimization methodology aims to select the best combination of sampled nodes and data offloading configuration to maximize FedL training accuracy while minimizing data processing and D2D communication resource consumption subject to realistic constraints on the network topology and device capabilities. Theoretical analysis of the D2D offloading subproblem leads to new FedL convergence bounds and an efficient sequential convex optimizer. Using these results, we develop a sampling methodology based on graph convolutional networks (GCNs) which learns the relationship between network attributes, sampled nodes, and D2D data offloading to maximize FedL accuracy. Through evaluation on popular datasets and real-world network measurements from our edge testbed, we find that our methodology outperforms popular device sampling methodologies from literature in terms of ML model performance, data processing overhead, and energy consumption., Comment: Published in IEEE/ACM Transactions on Networking. arXiv admin note: substantial text overlap with arXiv:2101.00787

The network edge's role in Artificial Intelligence (AI) inference processing is rapidly expanding, driven by a plethora of applications seeking computational advantages. These applications strive for data-driven efficiency, leveraging robust AI capabilities and prioritizing real-time responsiveness. However, as demand grows, so does system complexity. The proliferation of AI inference accelerators showcases innovation but also underscores challenges, particularly the varied software and hardware configurations of these devices. This diversity, while advantageous for certain tasks, introduces hurdles in device integration and coordination. In this paper, our objectives are three-fold. Firstly, we outline the requirements and components of a framework that accommodates hardware diversity. Next, we assess the impact of device heterogeneity on AI inference performance, identifying strategies to optimize outcomes without compromising service quality. Lastly, we shed light on the prevailing challenges and opportunities in this domain, offering insights for both the research community and industry stakeholders., Comment: This paper has been accepted for publication in the proceedings of the IEEE International Workshop on Computer Aided Modeling and Design of Communication Links and Networks 2023 (IEEE CAMAD 2023)

In this paper we combine the Early Science radio continuum data from the MeerKAT International GHz Tiered Extragalactic Exploration (MIGHTEE) Survey, with optical and near-infrared data and release the cross-matched catalogues. The radio data used in this work covers $0.86$ deg$^2$ of the COSMOS field, reaches a thermal noise of $1.7$ $\mu$Jy/beam and contains $6102$ radio components. We visually inspect and cross-match the radio sample with optical and near-infrared data from the Hyper Suprime-Cam (HSC) and UltraVISTA surveys. This allows the properties of active galactic nuclei and star-forming populations of galaxies to be probed out to $z \approx 5$. Additionally, we use the likelihood ratio method to automatically cross-match the radio and optical catalogues and compare this to the visually cross-matched catalogue. We find that 94 per cent of our radio source catalogue can be matched with this method, with a reliability of $95$ per cent. We proceed to show that visual classification will still remain an essential process for the cross-matching of complex and extended radio sources. In the near future, the MIGHTEE survey will be expanded in area to cover a total of $\sim$20~deg$^2$; thus the combination of automated and visual identification will be critical. We compare redshift distribution of SFG and AGN to the SKADS and T-RECS simulations and find more AGN than predicted at $z \sim 1$., Comment: 15 pages, 15 figures. Accepted for publication in MNRAS

We present the first wide area (2.5 x 2.5 deg^2) LOFAR High Band Antenna image at a resolution of 1.2'' x 2'' with a median noise of approximately 80 microJy per beam. It was made from an 8-hour International LOFAR Telescope (ILT) observation of the ELAIS-N1 field at frequencies ranging from 120 to 168 MHz with the most up-to-date ILT imaging methods. This intermediate resolution falls between the highest possible resolution (0.3'') achievable by using all International LOFAR Telescope (ILT) baselines and the standard 6-arcsecond resolution in the LoTSS (LOFAR Two-meter Sky Survey) image products utilising the LOFAR Dutch baselines only. This is the first demonstration of the feasibility of imaging using the ILT at a resolution of around 1'', which provides unique information on source morphology at scales that fall below the surface brightness limits at higher resolutions. The total calibration and imaging computational time is approximately 52,000 core hours, nearly 5 times more than required to produce a 6'' resolution image. We also present a radio source catalogue containing 2263 sources detected over the 2.5 x 2.5 deg^2 image of the ELAIS-N1 field, with a peak intensity threshold of 5.5 sigma. The catalogue has been cross-matched with the LoTSS deep ELAIS-N1 field radio catalogue, and its flux density and positional accuracy have been investigated and corrected accordingly. We find that approximately 80% of sources which we expect to be detectable based on their peak brightness in the LoTSS 6'' resolution image are detected in this image, which is approximately a factor of two higher than for 0.3'' resolution imaging in the Lockman Hole, implying there is a wealth of information on these intermediate scales., Comment: 20 pages, 19 figures, revised manuscript submitted to A&A

This paper proposes the neural publish/subscribe paradigm, a novel approach to orchestrating AI workflows in large-scale distributed AI systems in the computing continuum. Traditional centralized broker methodologies are increasingly struggling with managing the data surge resulting from the proliferation of 5G systems, connected devices, and ultra-reliable applications. Moreover, the advent of AI-powered applications, particularly those leveraging advanced neural network architectures, necessitates a new approach to orchestrate and schedule AI processes within the computing continuum. In response, the neural pub/sub paradigm aims to overcome these limitations by efficiently managing training, fine-tuning and inference workflows, improving distributed computation, facilitating dynamic resource allocation, and enhancing system resilience across the computing continuum. We explore this new paradigm through various design patterns, use cases, and discuss open research questions for further exploration.

About ten years ago, a paper proposed the first integer linear programming formulation for the constrained two-dimensional guillotine cutting problem (with unlimited cutting stages). Since, six other formulations followed, five of them in the last two years. This spike of interest gave no opportunity for a comprehensive comparison between the formulations. We review each formulation and compare their empirical results over instance datasets of the literature. We adapt most formulations to allow for piece rotation. The possibility of adaptation was already predicted but not realized by the prior work. The results show the dominance of pseudo-polynomial formulations until the point instances become intractable by them, while more compact formulations keep achieving good primal solutions. Our study also reveals a small but consistent advantage of the Gurobi solver over the CPLEX solver in our context; that the choice of solver hardly benefits one formulation over another; and a mistake in the generation of the T instances, which should have the same optima with or without guillotine cuts. Our study also proposes hybridising the most recent formulation with a prior formulation for a restricted version of the problem. The hybridisations show a reduction of about 20% of the branch-and-bound time thanks to the symmetries broken by the hybridisation., Comment: 23 pages, 7 tables, 3 figures

We present an International LOFAR Telescope sub-arcsecond resolution image of the nearby galaxy M 51 with a beam size of 0.436" x 0.366" and rms of 46 $\mu$Jy. We compare this image with an European VLBI Network study of M 51, and discuss the supernovae in this galaxy, which have not yet been probed at these low radio frequencies. We find a flux density of 0.97 mJy for SN 2011dh in the ILT image, which is about five times smaller than the flux density reported by the LOFAR Twometre Sky Survey at 6" resolution using the same dataset without the international stations. This difference makes evident the need for LOFAR international baselines to reliably obtain flux density measurements of compact objects in nearby galaxies. Our LOFAR flux density measurement of SN 2011dh directly translates into fitting the radio light curves for the supernova and constraining massloss rates of progenitor star. We do not detect two other supernovae in the same galaxy, SN 1994I and SN 2005cs, and our observations place limits on the evolution of both supernovae at radio wavelengths. We also discuss the radio emission from the centre of M 51, in which we detect the Active Galactic Nucleus and other parts of the nuclear emission in the galaxy, and a possible detection of Component N. We discuss a few other sources, including the detection of a High mass X-ray Binary not detected by LoTSS, but with a flux density in the ILT image that matches well with higher frequency catalogues., Comment: Accepted for publication in the Astrophysical Journal. 16 pages, 5 figures

Radio-mode feedback is a key ingredient in galaxy formation and evolution models, required to reproduce the observed properties of massive galaxies in the local Universe. We study the cosmic evolution of radio-AGN feedback out to $z\sim2.5$ using a sample of 9485 radio-excess AGN. We combine the evolving radio luminosity functions with a radio luminosity scaling relationship to estimate AGN jet kinetic powers and derive the cosmic evolution of the kinetic luminosity density, $\Omega_{\rm{kin}}$ (i.e. the volume-averaged heating output). Compared to all radio-AGN, low-excitation radio galaxies (LERGs) dominate the feedback activity out to $z\sim2.5$, with both these populations showing a constant heating output of $\Omega_{\rm{kin}} \approx 4-5 \times 10^{32}\,\rm{W\,Mpc^{-3}}$ across $0.5 < z < 2.5$. We compare our observations to predictions from semi-analytical and hydrodynamical simulations, which broadly match the observed evolution in $\Omega_{\rm{kin}}$, although their absolute normalisation varies. Comparison to the Semi-Analytic Galaxy Evolution (SAGE) model suggests that radio-AGN may provide sufficient heating to offset radiative cooling losses, providing evidence for a self-regulated AGN feedback cycle. We integrate the kinetic luminosity density across cosmic time to obtain the kinetic energy density output from AGN jets throughout cosmic history to be $\sim 10^{50}\,\rm{J\,Mpc^{-3}}$. Compared to AGN winds, the kinetic energy density from AGN jets dominates the energy budget at $z \lesssim 2$; this suggests that AGN jets play an important role in AGN feedback across most of cosmic history., Comment: 15 pages, 6 figures, accepted for publication in MNRAS

It has long been known that luminous, ultra-steep spectrum radio sources are preferentially associated with massive galaxies at high redshifts. Here we describe a pilot project directed at such objects, to demonstrate the feasibility and importance of using LOFAR to study the most distant forming massive galaxies and protoclusters. We have successfully imaged four high-redshift ($z>2$) high-luminosity radio galaxies with sub-arcsecond resolution, at 144 MHz, using the International LOFAR Telescope (ILT). Our targets were 4C 41.17 ($z=3.8$), the "Anthill", B2 0902+34 ($z=3.4$), 4C 34.34 ($z=2.4$) and 4C 43.15 ($z=2.5$). Their low-frequency morphologies and the spatial distributions of their low-frequency spectral indices have been mapped, and compared with available optical, infrared, and X-ray images. Both for the Anthill at $z = 3.8$ and B2 0902+34 at $z=3.4$, the location of the steepest radio emission coincides with the Ly$\alpha$ emitting ionized gas halo. Our pilot project demonstrates that, because of its outstanding sensitivity and high angular resolution at low frequencies, the ILT is a unique facility for studying the co-evolution and interaction of massive galaxies, galaxy clusters, and supermassive black holes in the early Universe., Comment: 10 pages, 10 figures, 2 tables

Debiasing methods that seek to mitigate the tendency of Language Models (LMs) to occasionally output toxic or inappropriate text have recently gained traction. In this paper, we propose a standardized protocol which distinguishes methods that yield not only desirable results, but are also consistent with their mechanisms and specifications. For example, we ask, given a debiasing method that is developed to reduce toxicity in LMs, if the definition of toxicity used by the debiasing method is reversed, would the debiasing results also be reversed? We used such considerations to devise three criteria for our new protocol: Specification Polarity, Specification Importance, and Domain Transferability. As a case study, we apply our protocol to a popular debiasing method, Self-Debiasing, and compare it to one we propose, called Instructive Debiasing, and demonstrate that consistency is as important an aspect to debiasing viability as is simply a desirable result. We show that our protocol provides essential insights into the generalizability and interpretability of debiasing methods that may otherwise go overlooked., Comment: 9 pages (excluding references), accepted at ACL Findings 2023