Search

The global banking system has faced increasing challenges in combating money laundering, necessitating advanced methods for detecting suspicious transactions. Anti-money laundering (or AML) approaches have often relied on predefined thresholds and machine learning algorithms using flagged transaction data, which are limited by the availability and accuracy of existing datasets. In this paper, we introduce a novel algorithm that leverages network analysis to detect potential money laundering activities within large-scale transaction data. Utilizing an anonymized transactional dataset from Co\"operatieve Rabobank U.A., our method combines community detection via the Louvain algorithm and small cycle detection to identify suspicious transaction patterns below the regulatory reporting thresholds. Our approach successfully identifies cycles of transactions that may indicate layering steps in money laundering, providing a valuable tool for financial institutions to enhance their AML efforts. The results suggest the efficacy of our algorithm in pinpointing potentially illicit activities that evade current detection methods.

Most datacenter transport protocols traditionally depend on in-order packet delivery, a legacy design choice that prioritizes simplicity. However, technological advancements, such as RDMA, now enable the relaxation of this requirement, allowing for more efficient utilization of modern datacenter topologies like FatTree and Dragonfly. With the growing prevalence of AI/ML workloads, the demand for improved link utilization has intensified, creating challenges for single-path load balancers due to problems like ECMP collisions. In this paper, we present ARCANE, a novel, adaptive per-packet traffic load-balancing algorithm designed to work seamlessly with existing congestion control mechanisms. ARCANE dynamically routes packets to bypass congested areas and network failures, all while maintaining a lightweight footprint with minimal state requirements. Our evaluation shows that ARCANE delivers significant performance gains over traditional load-balancing methods, including packet spraying and other advanced solutions, substantially enhancing both performance and link utilization in modern datacenter networks.

We investigate the lazy burning process for Latin squares by studying their associated hypergraphs. In lazy burning, a set of vertices in a hypergraph is initially burned, and that burning spreads to neighboring vertices over time via a specified propagation rule. The lazy burning number is the minimum number of initially burned vertices that eventually burns all vertices. The hypergraphs associated with Latin squares include the $n$-uniform hypergraph, whose vertices and hyperedges correspond to the entries and lines (that is, sets of rows, columns, or symbols) of the Latin square, respectively, and the $3$-uniform hypergraph, which has vertices corresponding to the lines of the Latin square and hyperedges induced by its entries. Using sequences of vertices that together form a vertex cover, we show that for a Latin square of order $n$, the lazy burning number of its $n$-uniform hypergraph is bounded below by $n^2-3n+3$ and above by $n^2-3n+2 + \lfloor \log_2 n \rfloor.$ These bounds are shown to be tight using cyclic Latin squares and powers of intercalates. For the $3$-uniform hypergraph case, we show that the lazy burning number of Latin squares is one plus its shortest connected chain of subsquares. We determine the lazy burning number of Latin square hypergraphs derived from finitely generated groups. We finish with open problems.

We exploit the DustPedia sample of galaxies within approximately 40 Mpc, selecting 388 sources, to investigate the correlations between IR luminosity (L$_{\rm IR}$), the star formation rate (SFR), and the CO(1-0) luminosity (L$_{\rm CO}$) down to much lower luminosities than reached by previous analyses. We find a sub-linear dependence of the SFR on L$_{\rm IR}$. Below $\log(\hbox{L}_{\rm IR}/\hbox{L}_\odot)\simeq 10$ or $\hbox{SFR}\simeq 1\,\hbox{M}_\odot\,\hbox{yr}^{-1}$, the SFR/L$_{\rm IR}$ ratio substantially exceeds the standard ratio for dust-enshrouded star formation, and the difference increases with decreasing L$_{\rm IR}$ values. This implies that the effect of unobscured star formation overcomes that of dust heating by old stars, at variance with results based on the $\textit{Planck}$ ERCSC galaxy sample. We also find that the relations between the L$_{\rm CO}$ and L$_{\rm IR}$ or the SFR are consistent with those obtained at much higher luminosities., Comment: 9 pages, 5 figures, accepted for publication in Galaxies

We argue that the difference in infrared-to-radio luminosity ratio between local and high-redshift star-forming galaxies reflects {the alternative physical conditions} -- including magnetic field configurations -- of the dominant population of star-forming galaxies in different redshift ranges. We define three galactic types, based on our reference model, with reference to ages of stellar populations. ``Normal'' late-type galaxies dominate the star formation in the nearby Universe; ``starburst'' galaxies take over at higher redshifts, up to z ~1.5; while ``protospheroidal'' galaxies dominate at high redshift. A reanalysis of data from the COSMOS field combined with literature results shows that, for each population, the data are consistent with an almost redshift-independent mean value of the parameter q_IR, which quantifies the infrared-radio correlation. However, we find a hint of an upturn of the mean q_IR at z>~3.5 consistent with the predicted dimming of synchrotron emission due to cooling of relativistic electrons by inverse Compton scattering off the cosmic microwave background. The typical stellar masses increase from normal, to starburst, and to protospheroidal galaxies, accounting for the reported dependence of the mean q_IR on stellar mass. Higher values of q_IR found for high-z strongly lensed dusty galaxies selected at 500 micron might be explained by differential magnification., Comment: 9 pages, 4 figures, accepted for publication in A&A

The use of skeletal data allows deep learning models to perform action recognition efficiently and effectively. Herein, we believe that exploring this problem within the context of Continual Learning is crucial. While numerous studies focus on skeleton-based action recognition from a traditional offline perspective, only a handful venture into online approaches. In this respect, we introduce CHARON (Continual Human Action Recognition On skeletoNs), which maintains consistent performance while operating within an efficient framework. Through techniques like uniform sampling, interpolation, and a memory-efficient training stage based on masking, we achieve improved recognition accuracy while minimizing computational overhead. Our experiments on Split NTU-60 and the proposed Split NTU-120 datasets demonstrate that CHARON sets a new benchmark in this domain. The code is available at https://github.com/Sperimental3/CHARON., Comment: Accepted at ICPR 2024

We give exact values and bounds on the isoperimetric peak of complete trees, improving on known results. For the complete $q$-ary tree of depth $d$, if $q\ge 5$, then we find that the isoperimetric peak equals $d$, completing an open problem. In the case that $q$ is 3 or 4, we determine the value up to three values, and in the case $q=2$, up to a logarithmic additive factor. Our proofs use novel compression techniques, including left, down, and aeolian compressions. We apply our results to show that the vertex separation number and the isoperimetric peak of a graph may be arbitrarily far apart as a function of the order of the graph and give new bounds on the pathwidth and pursuit-evasion parameters on complete trees.

We present a model of radio continuum emission associated with star formation (SF) and active galactic nuclei (AGN) implemented in the Shark semi-analytic model of galaxy formation. SF emission includes free-free and synchrotron emission, which depend on the free-electron density and the rate of core-collapse supernovae with a minor contribution from supernova remnants, respectively. AGN emission is modelled based on the jet production rate, which depends on the black hole mass, accretion rate and spin, and includes synchrotron self-absorption. Shark reproduces radio luminosity functions (RLFs) at 1.4 GHz and 150 MHz for 0 $\leq$ z $\leq$ 4, and scaling relations between radio luminosity, star formation rate and infrared luminosity of galaxies in the local and distant universe in good agreement with observations. The model also reproduces observed number counts of radio sources from 150 MHz to 8.4 GHz to within a factor of two on average, though larger discrepancies are seen at the very bright fluxes at higher frequencies. We use this model to understand how the radio continuum emission from radio-quiet AGNs can affect the measured RLFs of galaxies. We find current methods to exclude AGNs from observational samples result in large fractions of radio-quiet AGNs contaminating the "star-forming galaxies" selection and a brighter end to the resulting RLFs. We investigate how this effects the infrared-radio correlation (IRRC) and show that AGN contamination can lead to evolution of the IRRC with redshift. Without this contamination our model predicts a redshift- and stellar mass-independent IRRC, except at the dwarf-galaxy regime., Comment: Accepted for publication in MNRAS. 17 pages, 7 figures

In this paper, we introduce Selective-distillation for Class and Architecture-agnostic unleaRning (SCAR), a novel approximate unlearning method. SCAR efficiently eliminates specific information while preserving the model's test accuracy without using a retain set, which is a key component in state-of-the-art approximate unlearning algorithms. Our approach utilizes a modified Mahalanobis distance to guide the unlearning of the feature vectors of the instances to be forgotten, aligning them to the nearest wrong class distribution. Moreover, we propose a distillation-trick mechanism that distills the knowledge of the original model into the unlearning model with out-of-distribution images for retaining the original model's test performance without using any retain set. Importantly, we propose a self-forget version of SCAR that unlearns without having access to the forget set. We experimentally verified the effectiveness of our method, on three public datasets, comparing it with state-of-the-art methods. Our method obtains performance higher than methods that operate without the retain set and comparable w.r.t the best methods that rely on the retain set.

The increasing demand of machine learning (ML) workloads in datacenters places significant stress on current congestion control (CC) algorithms, many of which struggle to maintain performance at scale. These workloads generate bursty, synchronized traffic that requires both rapid response and fairness across flows. Unfortunately, existing CC algorithms that rely heavily on delay as a primary congestion signal often fail to react quickly enough and do not consistently ensure fairness. In this paper, we propose FASTFLOW, a streamlined sender-based CC algorithm that integrates delay, ECN signals, and optional packet trimming to achieve precise, real-time adjustments to congestion windows. Central to FASTFLOW is the QuickAdapt mechanism, which provides accurate bandwidth estimation at the receiver, enabling faster reactions to network conditions. We also show that FASTFLOW can effectively enhance receiver-based algorithms such as EQDS by improving their ability to manage in-network congestion. Our evaluation reveals that FASTFLOW outperforms cutting-edge solutions, including EQDS, Swift, BBR, and MPRDMA, delivering up to 50% performance improvements in modern datacenter networks.

We study the Localization game on locally finite graphs trees, where each of the countably many vertices have finite degree. In contrast to the finite case, we construct a locally finite tree with localization number $n$ for any choice of positive integer $n$. Our examples have uncountably many ends, and we show that this is necessary by proving that locally finite trees with finitely or countably many ends have localization number at most 2. Finally, as is the case for finite graphs, we prove that any locally finite graph contains a subdivision where one cop can capture the robber.

This study aims to investigate the variability of exposure levels among road users generated in a realistic urban scenario by Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication technologies operating at 5.9 GHz. The exposure levels were evaluated in terms of whole-body Specific Absorption Rate (wbSAR) [W/kg] in three different human models, ranging from children to adults. We calculated the electromagnetic field exposure level generated by V2V and V2I using raytracing and we assessed wbSAR resulting in urban exposure scenarios with an increasing number of transmitting antennas. Whole-body SAR was generally very low, on the order of 10^-4 W/kg. The maximum wbSAR, of 4.9x10^-4 W/kg, was obtained in the worst-case exposure condition comprising more than one transmitting vehicle and was found in the adult model for a distance within 10 m from the transmitting cars. We found that the height of the human model highly impacted the exposure level. Namely, the child (which is the shortest human model) was generally much less exposed than adults. All the wbSAR values found by varying the number of transmitting antennas, the distance of the road user from the antennas, and the type of human model (adult vs. child) were very well below the limits set by the ICNIRP and IEEE guidelines of 0.08 W/kg for human exposure in the 100 kHz - 300 GHz range.

Quantum communication promises unprecedented communication capabilities enabled by the transmission of quantum states of light. However, current implementations face severe limitations in communication distance due to photon loss. Silicon carbide (SiC) defects have emerged as a promising quantum device platform, offering strong optical transitions, long spin coherence lifetimes and the opportunity for integration with semiconductor devices. Some defects with optical transitions in the telecom range have been identified, allowing to interface with fiber networks without the need for wavelength conversion. These unique properties make SiC an attractive platform for the implementation of quantum nodes for quantum communication networks. We provide an overview of the most prominent defects in SiC and their implementation in spin-photon interfaces. Furthermore, we model a memory-enhanced quantum communication protocol in order to extract the parameters required to surpass a direct point-to-point link performance. Based on these insights, we summarize the key steps required towards the deployment of SiC devices in large-scale quantum communication networks., Comment: 20 pages, 8 figures

The allreduce collective operation accounts for a significant fraction of the runtime of workloads running on distributed systems. One factor determining its performance is the distance between communicating nodes, especially on networks like torus, where a higher distance implies multiple messages being forwarded on the same link, thus reducing the allreduce bandwidth. Torus networks are widely used on systems optimized for machine learning workloads (e.g., Google TPUs and Amazon Trainium devices), as well as on some of the Top500 supercomputers. To improve allreduce performance on torus networks we introduce Swing, a new algorithm that keeps a low distance between communicating nodes by swinging between torus directions. Our analysis and experimental evaluation show that Swing outperforms by up to 3x existing allreduce algorithms for vectors ranging from 32B to 128MiB, on different types of torus and torus-like topologies, regardless of their shape and size.

Counts of small subgraphs, or graphlet counts, are widely applicable to measure graph similarity. Computing graphlet counts can be computationally expensive and may pose obstacles in network analysis. We study the role of cliques in graphlet counts as a method for graph similarity in social networks. Higher-order clustering coefficients and the Pivoter algorithm for exact clique counts are employed

We introduce a new graph parameter called the cooling number, inspired by the spread of influence in networks and its predecessor, the burning number. The cooling number measures the speed of a slow-moving contagion in a graph; the lower the cooling number, the faster the contagion spreads. We provide tight bounds on the cooling number via a graph's order and diameter. Using isoperimetric results, we derive the cooling number of Cartesian grids. The cooling number is studied in graphs generated by the Iterated Local Transitivity model for social networks. We conclude with open problems.

The 3D folding of a mammalian gene can be studied by a polymer model, where the chromatin fibre is represented by a semiflexible polymer which interacts with multivalent proteins, representing complexes of DNA-binding transcription factors and RNA polymerases. This physical model leads to the natural emergence of clusters of proteins and binding sites, accompanied by the folding of chromatin into a set of topologies, each associated with a different network of loops. Here we combine numerics and analytics to first classify these networks and then find their relative importance or statistical weight, when the properties of the underlying polymer are those relevant to chromatin. Unlike polymer networks previously studied, our chromatin networks have finite average distances between successive binding sites, and this leads to giant differences between the weights of topologies with the same number of edges and nodes but different wiring. These weights strongly favour rosette-like structures with a local cloud of loops with respect to more complicated non-local topologies. Our results suggest that genes should overwhelmingly fold into a small fraction of all possible 3D topologies, which can be robustly characterised by the framework we propose here., Comment: 8 pages, 2 figures

Polymer physics models suggest that chromatin spontaneously folds into loop networks with transcription units (TUs), such as enhancers and promoters, as anchors. Here we use combinatoric arguments to enumerate the emergent chromatin loop networks, both in the case where TUs are labelled and where they are unlabelled. We then combine these mathematical results with those of computer simulations aimed at finding the inter-TU energy required to form a target loop network. We show that different topologies are vastly different in terms of both their combinatorial weight and energy of formation. We explain the latter result qualitatively by computing the topological weight of a given network -- i.e., its partition function in statistical mechanics language -- in the approximation where excluded volume interactions are neglected. Our results show that networks featuring local loops are statistically more likely with respect to networks including more non-local contacts. We suggest our classification of loop networks, together with our estimate of the combinatorial and topological weight of each network, will be relevant to catalogue 3D structures of chromatin fibres around eukaryotic genes, and to estimate their relative frequency in both simulations and experiments., Comment: 19 pages, 11 figures

The field of nanoscale magnetic resonance imaging (NanoMRI) was started 30 years ago. It was motivated by the desire to image single molecules and molecular assemblies, such as proteins and virus particles, with near-atomic spatial resolution and on a length scale of 100 nm. Over the years, the NanoMRI field has also expanded to include the goal of useful high-resolution nuclear magnetic resonance (NMR) spectroscopy of molecules under ambient conditions, including samples up to the micron-scale. The realization of these goals requires the development of spin detection techniques that are many orders of magnitude more sensitive than conventional NMR and MRI, capable of detecting and controlling nanoscale ensembles of spins. Over the years, a number of different technical approaches to NanoMRI have emerged, each possessing a distinct set of capabilities for basic and applied areas of science. The goal of this roadmap article is to report the current state of the art in NanoMRI technologies, outline the areas where they are poised to have impact, identify the challenges that lie ahead, and propose methods to meet these challenges. This roadmap also shows how developments in NanoMRI techniques can lead to breakthroughs in emerging quantum science and technology applications.

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

The recent surge of pervasive devices that generate dynamic data streams has underscored the necessity for learning systems to adapt continually to data distributional shifts. To tackle this challenge, the research community has put forth a spectrum of methodologies, including the demanding pursuit of class-incremental learning without replay data. In this study, we present MIND, a parameter isolation method that aims to significantly enhance the performance of replay-free solutions and achieve state-of-the-art results on several widely studied datasets. Our approach introduces two main contributions: two alternative distillation procedures that significantly improve the efficiency of MIND increasing the accumulated knowledge of each sub-network, and the optimization of the BachNorm layers across tasks inside the sub-networks. Overall, MIND outperforms all the state-of-the-art methods for rehearsal-free Class-Incremental learning (with an increment in classification accuracy of approx. +6% on CIFAR-100/10 and +10% on TinyImageNet/10) reaching up to approx. +40% accuracy in Domain-Incremental scenarios. Moreover, we ablated each contribution to demonstrate its impact on performance improvement. Our results showcase the superior performance of MIND indicating its potential for addressing the challenges posed by Class-incremental and Domain-Incremental learning in resource-constrained environments., Comment: Accepted at the 38th AAAI Conference on Artificial Intelligence

Machine Unlearning is rising as a new field, driven by the pressing necessity of ensuring privacy in modern artificial intelligence models. This technique primarily aims to eradicate any residual influence of a specific subset of data from the knowledge acquired by a neural model during its training. This work introduces a novel unlearning algorithm, denoted as Distance-based Unlearning via Centroid Kinematics (DUCK), which employs metric learning to guide the removal of samples matching the nearest incorrect centroid in the embedding space. Evaluation of the algorithm's performance is conducted across various benchmark datasets in two distinct scenarios, class removal, and homogeneous sampling removal, obtaining state-of-the-art performance. We also introduce a novel metric, called Adaptive Unlearning Score (AUS), encompassing not only the efficacy of the unlearning process in forgetting target data but also quantifying the performance loss relative to the original model. Additionally, we conducted a thorough investigation of the unlearning mechanism in DUCK, examining its impact on the organization of the feature space and employing explainable AI techniques for deeper insights.

We study a variant of the Localization game in which the cops have limited visibility, along with the corresponding optimization parameter, the $k$-visibility localization number $\zeta_k$, where $k$ is a non-negative integer. We give bounds on $k$-visibility localization numbers related to domination, maximum degree, and isoperimetric inequalities. For all $k$, we give a family of trees with unbounded $\zeta_k$ values. Extending results known for the localization number, we show that for $k\geq 2$, every tree contains a subdivision with $\zeta_k = 1$. For many $n$, we give the exact value of $\zeta_k$ for the $n \times n$ Cartesian grid graphs, with the remaining cases being one of two values as long as $n$ is sufficiently large. These examples also illustrate that $\zeta_i \neq \zeta_j$ for all distinct choices of $i$ and $j.$

We present JWST/MIRI/MRS observations of an infrared luminous disk galaxy, FLS1, at z=0.54. With a lookback time of 5 Gyr, FLS1 is chronologically at the midpoint between the peak epoch of star formation and the present day. The MRS data provide maps of the atomic fine structure lines [Ar II]6.99 micron, [Ar III]8.99 micron, [Ne II]12.81 micron, and [Ne III]15.55 micron, polycyclic aromatic hydrocarbon (PAH) features at 3.3 micron, 6.2 micron, and 11.3 micron, and the warm molecular gas indicators H2S(5) and H2S(3); all these emission features are spatially resolved. We find that the PAH emission is more extended along the Northern side of the galaxy when compared to the well-studied star-formation tracer [Ne II]. The H2 rotational lines, which are shock indicators, are strongest and most extended on the Southern side of the galaxy. [Ar II] is the second brightest fine structure line detected in FLS1 and we show that it is a useful kinematic probe which can be detected with JWST out to z=3. Velocity maps of [Ar II] show a rotating disk with signs of turbulence. Our results provide an example of how spatially resolved mid-infrared spectroscopy can allow us to better understand the star formation and ISM conditions in a galaxy halfway back to the peak epoch of galaxy evolution., Comment: 10 pages, 5 figures, 1 table

DNA loop extrusion is arguably one of the most important players in genome organization. The precise mechanism by which loop extruding factors (LEFs) work is still unresolved and much debated. One of the major open questions in this field is how do LEFs establish and maintain unidirectional motion along DNA. In this paper, we use High-Speed AFM data to show that condensin hinge domain displays a structural, geometric constraint on the angle within which it can extend with respect to the DNA-bound domains. Using computer simulations, we then show that such a geometrical constraint results in a local symmetry breaking and is enough to rectify the extrusion process, yielding unidirectional loop extrusion along DNA. Our work highlights an overlooked geometric aspect of the loop extrusion process that may have a universal impact on SMC function across organisms.

A space-based far-infrared interferometer could work synergistically with the James Webb Space Telescope (JWST) and the Atacama Large Millimeter Array (ALMA) to revolutionize our understanding of the astrophysical processes leading to the formation of habitable planets and the co-evolution of galaxies and their central supermassive black holes. Key to these advances are measurements of water in its frozen and gaseous states, observations of astronomical objects in the spectral range where most of their light is emitted, and access to critical diagnostic spectral lines, all of which point to the need for a far-infrared observatory in space. The objects of interest - circumstellar disks and distant galaxies - typically appear in the sky at sub-arcsecond scales, which rendered all but a few of them unresolvable with the successful and now-defunct 3.5-m Herschel Space Observatory, the largest far-infrared telescope flown to date. A far-infrared interferometer with maximum baseline length in the tens of meters would match the angular resolution of JWST at 10x longer wavelengths and observe water ice and water-vapor emission, which ALMA can barely do through the Earth's atmosphere. Such a facility was conceived and studied two decades ago. Here we revisit the science case for a space-based far-infrared interferometer in the era of JWST and ALMA and summarize the measurement capabilities that will enable the interferometer to achieve a set of compelling scientific objectives. Common to all the science themes we consider is a need for sub-arcsecond image resolution., Comment: 11 pages, 12 figures. To be published in Proc. SPIE Conf. 12686 "Instruments: Scientific Returns and Conceptual Designs"

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

In this paper we extend the Tiered Radio Extragalactic Continuum Simulation (T-RECS) to include HI emission. The HI T-RECS model is based on the most recent HI mass function estimates, combined with prescriptions to convert HI mass to total integrated HI flux. It further models source size, morphology and kinematics, including rotational velocity and HI line width. The continuum T-RECS model is updated to improve the agreement with deeper number counts available at 150\,MHz. The model for star-forming galaxies (SFGs) is also modified according to the most recent indications of a star formation rate (SFR)--radio luminosity relation, which depends primarily on stellar mass rather than redshift. We further introduce prescriptions to associate an HI mass to the T-RECS radio continuum SFG and Active Galactic Nuclei (AGN) populations. This gives us a way to meaningfully associate counterparts between HI and continuum catalogues, thus building HI $\times$ continuum simulated observations. Clustering properties of the sources in both HI and continuum are reproduced by associating the galaxies to dark matter haloes of a cosmological simulation. We deliver a set of mock catalogues, as well as the code to produce them, which can be used for simulating observations and predicting results from radio surveys with existing and forthcoming radio facilities, such as the Square Kilometre Array (SKA), Comment: 18 pages, 10 figures, accepted by MNRAS

Source classifications, stellar masses and star formation rates are presented for 80,000 radio sources from the first data release of the Low Frequency Array Two-metre Sky Survey (LoTSS) Deep Fields, which represents the widest deep radio survey ever undertaken. Using deep multi-wavelength data spanning from the ultraviolet to the far-infrared, spectral energy distribution (SED) fitting is carried out for all of the LoTSS-Deep host galaxies using four different SED codes, two of which include modelling of the contributions from an active galactic nucleus (AGN). Comparing the results of the four codes, galaxies that host a radiative AGN are identified, and an optimised consensus estimate of the stellar mass and star-formation rate for each galaxy is derived. Those galaxies with an excess of radio emission over that expected from star formation are then identified, and the LoTSS-Deep sources are divided into four classes: star-forming galaxies, radio-quiet AGN, and radio-loud high-excitation and low-excitation AGN. Ninety-five per cent of the sources can be reliably classified, of which more than two-thirds are star-forming galaxies, ranging from normal galaxies in the nearby Universe to highly-starbursting systems at z>4. Star-forming galaxies become the dominant population below 150-MHz flux densities of about 1 mJy, accounting for 90 per cent of sources at a 150-MHz flux density of 100 microJy. Radio-quiet AGN comprise around 10 per cent of the overall population. Results are compared against the predictions of the SKADS and T-RECS radio sky simulations, and improvements to the simulations are suggested., Comment: Accepted for publication in MNRAS. Catalogues available at www.lofar-surveys.org/deepfields.html. 27 pages

Recent advancements have enabled human-robot collaboration through physical assistance and verbal guidance. However, limitations persist in coordinating robots' physical motions and speech in response to real-time changes in human behavior during collaborative contact tasks. We first derive principles from analyzing physical therapists' movements and speech during patient exercises. These principles are translated into control objectives to: 1) guide users through trajectories, 2) control motion and speech pace to align completion times with varying user cooperation, and 3) dynamically paraphrase speech along the trajectory. We then propose a Language Controller that synchronizes motion and speech, modulating both based on user cooperation. Experiments with 12 users show the Language Controller successfully aligns motion and speech compared to baselines. This provides a framework for fluent human-robot collaboration., Comment: Under review in ICRA 2024

Spin-active quantum emitters have emerged as a leading platform for quantum technologies. However, one of their major limitations is the large spread in optical emission frequencies, which typically extends over tens of GHz. Here, we investigate single V4+ vanadium centres in 4H-SiC, which feature telecom-wavelength emission and a coherent S=1/2 spin state. We perform spectroscopy on single emitters and report the observation of spin-dependent optical transitions, a key requirement for spin-photon interfaces. By engineering the isotopic composition of the SiC matrix, we reduce the inhomogeneous spectral distribution of different emitters down to 100 MHz, significantly smaller than any other single quantum emitter. Additionally, we tailor the dopant concentration to stabilise the telecom-wavelength V4+ charge state, thereby extending its lifetime by at least two orders of magnitude. These results bolster the prospects for single V emitters in SiC as material nodes in scalable telecom quantum networks., Comment: 29 pages, 20 figures, 2 tables