Search

Deep learning has been proven to be a powerful tool for addressing the most significant issues in cognitive radio networks, such as spectrum sensing, spectrum sharing, resource allocation, and security attacks. The utilization of deep learning techniques in cognitive radio networks can significantly enhance the network's capability to adapt to changing environments and improve the overall system's efficiency and reliability. As the demand for higher data rates and connectivity increases, B5G/6G wireless networks are expected to enable new services and applications significantly. Therefore, the significance of deep learning in addressing cognitive radio network challenges cannot be overstated. This review article provides valuable insights into potential solutions that can serve as a foundation for the development of future B5G/6G services. By leveraging the power of deep learning, cognitive radio networks can pave the way for the next generation of wireless networks capable of meeting the ever-increasing demands for higher data rates, improved reliability, and security., Comment: The article has been accepted for publication in "Journal of Network and Computer Applications" during October 2024

Spell-checkers are valuable tools that enhance communication by identifying misspelled words in written texts. Recent improvements in deep learning, and in particular in large language models, have opened new opportunities to improve traditional spell-checkers with new functionalities that not only assess spelling correctness but also the suitability of a word for a given context. In our work, we present and compare two new spell-checkers and evaluate them on synthetic, learner, and more general-domain Slovene datasets. The first spell-checker is a traditional, fast, word-based approach, based on a morphological lexicon with a significantly larger word list compared to existing spell-checkers. The second approach uses a language model trained on a large corpus with synthetically inserted errors. We present the training data construction strategies, which turn out to be a crucial component of neural spell-checkers. Further, the proposed neural model significantly outperforms all existing spell-checkers for Slovene in both precision and recall., Comment: Camera-ready version. Accepted to TSD 2024

As general-purpose tools, Large Language Models (LLMs) must often reason about everyday physical environments. In a question-and-answer capacity, understanding the interactions of physical objects may be necessary to give appropriate responses. Moreover, LLMs are increasingly used as reasoning engines in agentic systems, designing and controlling their action sequences. The vast majority of research has tackled this issue using static benchmarks, comprised of text or image-based questions about the physical world. However, these benchmarks do not capture the complexity and nuance of real-life physical processes. Here we advocate for a second, relatively unexplored, approach: 'embodying' the LLMs by granting them control of an agent within a 3D environment. We present the first embodied and cognitively meaningful evaluation of physical common-sense reasoning in LLMs. Our framework allows direct comparison of LLMs with other embodied agents, such as those based on Deep Reinforcement Learning, and human and non-human animals. We employ the Animal-AI (AAI) environment, a simulated 3D virtual laboratory, to study physical common-sense reasoning in LLMs. For this, we use the AAI Testbed, a suite of experiments that replicate laboratory studies with non-human animals, to study physical reasoning capabilities including distance estimation, tracking out-of-sight objects, and tool use. We demonstrate that state-of-the-art multi-modal models with no finetuning can complete this style of task, allowing meaningful comparison to the entrants of the 2019 Animal-AI Olympics competition and to human children. Our results show that LLMs are currently outperformed by human children on these tasks. We argue that this approach allows the study of physical reasoning using ecologically valid experiments drawn directly from cognitive science, improving the predictability and reliability of LLMs., Comment: 25 pages, 4 figures

We provide a solution for decoherence in spatial superpositions due to scattering/collision with air molecules. This result reproduces the short- and long-wavelength limits known in the literature. We compare the decoherence rate with several existing interpolations in the literature and evaluate the decoherence rate and experimental parameters when creating macroscopic quantum spatial superpositions (i.e., micron-size spheres). Finally, we consider the decoherence rate's time dependence while creating and closing the spatial superposition in an interferometer setup., Comment: 4 figures, 1 table

Persistent revivals recently observed in Rydberg atom simulators have challenged our understanding of thermalization and attracted much interest to the concept of quantum many-body scars (QMBSs). QMBSs are non-thermal highly excited eigenstates that coexist with typical eigenstates in the spectrum of many-body Hamiltonians, and have since been reported in multiple theoretical models, including the so-called PXP model, approximately realized by Rydberg simulators. At the same time, questions of how common QMBSs are and in what models they are physically realized remain open. In this Letter, we demonstrate that QMBSs exist in a broader family of models that includes and generalizes PXP to longer-range constraints and states with different periodicity. We show that in each model, multiple QMBS families can be found. Each of them relies on a different approximate $\mathfrak{su}(2)$ algebra, leading to oscillatory dynamics in all cases. However, in contrast to the PXP model, their observation requires launching dynamics from weakly entangled initial states rather than from a product state. QMBSs reported here may be experimentally probed using Rydberg atom simulator in the regime of longer-range Rydberg blockades., Comment: 8+12 pages, 6+8 figures

Two-dimensional (2D) heterostructures reveal novel physicochemical phenomena at different length scales, that are highly desirable for technological applications. We present a comprehensive density functional theory study of van der Waals (vdW) heterostructures constructed by stacking 2D TiO2 and 2D MoSSe monolayers to form TiO2-MoSSe heterojunction. The heterostructure formation is found to be exothermic, indicating stability. We find that by varying the atomic species at the interfaces the electronic structure can be considerably altered, due to the differences in charge transfer, arising from the inherent electronegativity of the atoms. We demonstrate that the heterostructures possess a type II or type III band alignment, depending on the atomic termination of MoSSe at the interface. The observed charge transfer occurs from MoSSe to TiO2. Our results suggest that Janus interface enables the tuning of electronic properties, providing understanding of the possible applications of the TiO2-MoSSe heterostructure., Comment: Main text: 25 pages, 5 figures. Supplementary information: 7 pages, 10 figures. Accepted manuscript for publication in ACS Omega

Python Testbed for Federated Learning Algorithms (PTB-FLA) is a simple FL framework targeting smart Internet of Things in edge systems that provides both generic centralized and decentralized FL algorithms, which implement the corresponding FL orchestration protocols that were formally verified using the process algebra CSP. This approach is appropriate for systems with stationary nodes but cannot be applied to systems with moving nodes. In this paper, we use celestial mechanics to model spacecraft movement, and timed automata (TA) to formalize and verify the centralized FL orchestration protocol, in two phases. In the first phase, we created a conventional TA model to prove traditional properties, namely deadlock freeness and termination. In the second phase, we created a stochastic TA model to prove timing correctness and to estimate termination probability., Comment: 4 pages, 5 figures, submitted to a conference

The sarcasm detection task in natural language processing tries to classify whether an utterance is sarcastic or not. It is related to sentiment analysis since it often inverts surface sentiment. Because sarcastic sentences are highly dependent on context, and they are often accompanied by various non-verbal cues, the task is challenging. Most of related work focuses on high-resourced languages like English. To build a sarcasm detection dataset for a less-resourced language, such as Slovenian, we leverage two modern techniques: a machine translation specific medium-size transformer model, and a very large generative language model. We explore the viability of translated datasets and how the size of a pretrained transformer affects its ability to detect sarcasm. We train ensembles of detection models and evaluate models' performance. The results show that larger models generally outperform smaller ones and that ensembling can slightly improve sarcasm detection performance. Our best ensemble approach achieves an $\text{F}_1$-score of 0.765 which is close to annotators' agreement in the source language., Comment: 4 pages, published in the Slovenian Conference on Artificial Intelligence

This paper addresses the Restricted Longest Common Subsequence (RLCS) problem, an extension of the well-known Longest Common Subsequence (LCS) problem. This problem has significant applications in bioinformatics, particularly for identifying similarities and discovering mutual patterns and important motifs among DNA, RNA, and protein sequences. Building on recent advancements in solving this problem through a general search framework, this paper introduces two novel heuristic approaches designed to enhance the search process by steering it towards promising regions in the search space. The first heuristic employs a probabilistic model to evaluate partial solutions during the search process. The second heuristic is based on a neural network model trained offline using a genetic algorithm. A key aspect of this approach is extracting problem-specific features of partial solutions and the complete problem instance. An effective hybrid method, referred to as the learning beam search, is developed by combining the trained neural network model with a beam search framework. An important contribution of this paper is found in the generation of real-world instances where scientific abstracts serve as input strings, and a set of frequently occurring academic words from the literature are used as restricted patterns. Comprehensive experimental evaluations demonstrate the effectiveness of the proposed approaches in solving the RLCS problem. Finally, an empirical explainability analysis is applied to the obtained results. In this way, key feature combinations and their respective contributions to the success or failure of the algorithms across different problem types are identified., Comment: 33 pages, 12 figures

The integrity of AI benchmarks is fundamental to accurately assess the capabilities of AI systems. The internal validity of these benchmarks - i.e., making sure they are free from confounding factors - is crucial for ensuring that they are measuring what they are designed to measure. In this paper, we explore a key issue related to internal validity: the possibility that AI systems can solve benchmarks in unintended ways, bypassing the capability being tested. This phenomenon, widely known in human and animal experiments, is often referred to as the 'Clever Hans' effect, where tasks are solved using spurious cues, often involving much simpler processes than those putatively assessed. Previous research suggests that language models can exhibit this behaviour as well. In several older Natural Language Processing (NLP) benchmarks, individual $n$-grams like "not" have been found to be highly predictive of the correct labels, and supervised NLP models have been shown to exploit these patterns. In this work, we investigate the extent to which simple $n$-grams extracted from benchmark instances can be combined to predict labels in modern multiple-choice benchmarks designed for LLMs, and whether LLMs might be using such $n$-gram patterns to solve these benchmarks. We show how simple classifiers trained on these $n$-grams can achieve high scores on several benchmarks, despite lacking the capabilities being tested. Additionally, we provide evidence that modern LLMs might be using these superficial patterns to solve benchmarks. This suggests that the internal validity of these benchmarks may be compromised and caution should be exercised when interpreting LLM performance results on them.

We study the representations of a class of non-commutative polynomial algebras truncated at degree 3, with one additional relation. We determine the irreducible components of their varieties of representations. We do this by showing that the subvarieties with fixed radical or socle layering are irreducible. We then compare the coverings we get using radical and socle layerings to determine the components., Comment: 21 pages, 1 figure

Substitutional doping with transition metals is carried out in the Lepidocrocite phase - the stable monolayer geometry of TiO$_2$, using density functional theory (DFT) methods. The doping is carried out at the differently coordinated O atom cites, producing Janus monolayer geometries. Our results indicate that key fundamental properties for photocatalysis can be tuned via doping. Monolayers doped with Ag, Au, Pd and Pt are thermodynamically stable, amongst all considered doping possibilities, as evident from phonon band structure calculations. Electronic structure of the Janus monolayers alters significantly, compared to pristine TiO$_2$, owing to the emergence of mid-gap states. Reduced band gap arises from upward shift of the valence band, suggesting enhanced visible-light response. Dopant atoms also introduce excess electrons in TiO$_2$ monolayers, which are found to localize at a single Ti site. This induces ferromagnetism in the doped monolayers. Furthermore, charge separation between TiO$_2$ and noble metal dopants is observed which is a key parameter in influencing the selectivity and activity of photocatalytic materials. Compared to the pristine TiO$_2$ monolayer, the Janus structure can promote water adsorption, and the Janus monolayers exhibit significantly improved activity in the hydrogen evolution reaction. These findings suggest that engineering a novel Janus TiO$_2$-based monolayer with a noble metal layer on the other surface can offer a potential approach to improve photocatalytic performance over pristine TiO$_2$., Comment: Main text: pages 20, figures 7. Supplementary information: pages 12, figures 12. Submitted for publication in Catalysis Today

Grid energy storage can help to balance supply and demand, but its financial viability and operational carbon emissions impact is poorly understood because of the complexity of grid constraints and market outcomes. We analyse the impact of several technologies (Li-ion and flow batteries, pumped hydro, hydrogen) on Great Britain balancing mechanism, the main market for supply-demand balancing and congestion management. We find that, for many locations and technologies, financially optimal operation of storage for balancing can result in higher carbon emissions. For example, the extra emissions associated with a 1 MW 2-hour duration Li-ion battery in winter vary between +230 to -71 kgCO2/h. Although storage enable higher usage of renewables, it can also unlock additional demand leading to greater use of gas. In addition, balancing services alone are presently insufficient for financial viability of storage projects. This work highlights the need for market reform aligning financial incentives with environmental impacts.

Mutual information provides a powerful, general-purpose metric for quantifying the amount of shared information between variables. Estimating normalized mutual information using a k-Nearest Neighbor (k-NN) based approach involves the calculation of the scaling-invariant k-NN radius. Calculation of the radius suffers from numerical overflow when the joint dimensionality of the data becomes high, typically in the range of several hundred dimensions. To address this issue, we propose a logarithmic transformation technique that improves the numerical stability of the radius calculation in high-dimensional spaces. By applying the proposed transformation during the calculation of the radius, numerical overflow is avoided, and precision is maintained. Proposed transformation is validated through both theoretical analysis and empirical evaluation, demonstrating its ability to stabilize the calculation without compromizing the precision of the results., Comment: 4+1 pages, 2 figures, 20 equations

Integrability is an exceptional property believed to hold only for systems with fine-tuned parameters. Contrary, we explicitly show that in homogeneous nearest-neighbor qubit circuits with a U(1) symmetry, i.e., circuits that repeatedly apply the same magnetization-conserving two-qubit gate, this is not the case. There, integrability is generic: all such brickwall qubit circuits are integrable, even with a randomly selected gate. We identify two phases with different conservation laws, transport properties, and strong zero edge modes. Experimentally important is the fact that varying any one of the parameters in the generic U(1) gate, one will typically cross the critical manifold that separates the two phases. Finally, we report on an unconventional time-reversal symmetry causing the system with open boundary conditions to be in the orthogonal class, while the one with periodic boundary conditions is in the unitary class., Comment: 8 pages

Large language models (LLMs) are a basic infrastructure for modern natural language processing. Many commercial and open-source LLMs exist for English, e.g., ChatGPT, Llama, Falcon, and Mistral. As these models are trained on mostly English texts, their fluency and knowledge of low-resource languages and societies are superficial. We present the development of large generative language models for a less-resourced language. GaMS 1B - Generative Model for Slovene with 1 billion parameters was created by continuing pretraining of the existing English OPT model. We developed a new tokenizer adapted to Slovene, Croatian, and English languages and used embedding initialization methods FOCUS and WECHSEL to transfer the embeddings from the English OPT model. We evaluate our models on several classification datasets from the Slovene suite of benchmarks and generative sentence simplification task SENTA. We only used a few-shot in-context learning of our models, which are not yet instruction-tuned. For classification tasks, in this mode, the generative models lag behind the existing Slovene BERT-type models fine-tuned for specific tasks. On a sentence simplification task, the GaMS models achieve comparable or better performance than the GPT-3.5-Turbo model.

Implicit Neural Representations (INRs) have recently gained attention as a powerful approach for continuously representing signals such as images, videos, and 3D shapes using multilayer perceptrons (MLPs). However, MLPs are known to exhibit a low-frequency bias, limiting their ability to capture high-frequency details accurately. This limitation is typically addressed by incorporating high-frequency input embeddings or specialized activation layers. In this work, we demonstrate that these embeddings and activations are often configured with hyperparameters that perform well on average but are suboptimal for specific input signals under consideration, necessitating a costly grid search to identify optimal settings. Our key observation is that the initial frequency spectrum of an untrained model's output correlates strongly with the model's eventual performance on a given target signal. Leveraging this insight, we propose frequency shifting (or FreSh), a method that selects embedding hyperparameters to align the frequency spectrum of the model's initial output with that of the target signal. We show that this simple initialization technique improves performance across various neural representation methods and tasks, achieving results comparable to extensive hyperparameter sweeps but with only marginal computational overhead compared to training a single model with default hyperparameters., Comment: Code at https://github.com/gmum/FreSh/

We study inhomogeneous non-strictly hyperbolic systems of two equations, which are a formal generalization of the transformed one-dimensional Euler-Poisson equations. For such systems, a complete classification of the behavior of the solution is carried out depending on the right side. In particular, criteria for the formation of singularities in the solution of the Cauchy problem in terms of initial data are found. Also we determine the domains of attraction of equilibria of the extended system for derivatives. We prove the existence of solutions in the form of simple waves. The results obtained are applied to study the main model cases of the Euler-Poisson equations.

In this paper, we study Artinian and Noetherian properties in vector lattices and provide a concrete representation of these spaces. Furthermore, we describe for which Archimedean uniformly complete vector lattices every decreasing sequence of prime ideals is stationary (a property that we refer to as prime Artinian). We also completely characterize the prime ideals in vector lattices of continuous root functions and piecewise polynomials. This is a useful space for studying how having decreasing stationary sequences of prime ideals does not imply having increasing stationary sequences of prime ideals, and vice versa.

In this paper, we propose a novel end-to-end relightable neural inverse rendering system that achieves high-quality reconstruction of geometry and material properties, thus enabling high-quality relighting. The cornerstone of our method is a two-stage approach for learning a better factorization of scene parameters. In the first stage, we develop a reflection-aware radiance field using a neural signed distance field (SDF) as the geometry representation and deploy an MLP (multilayer perceptron) to estimate indirect illumination. In the second stage, we introduce a novel information-sharing network structure to jointly learn the radiance field and the physically based factorization of the scene. For the physically based factorization, to reduce the noise caused by Monte Carlo sampling, we apply a split-sum approximation with a simplified Disney BRDF and cube mipmap as the environment light representation. In the relighting phase, to enhance the quality of indirect illumination, we propose a second split-sum algorithm to trace secondary rays under the split-sum rendering framework. Furthermore, there is no dataset or protocol available to quantitatively evaluate the inverse rendering performance for glossy objects. To assess the quality of material reconstruction and relighting, we have created a new dataset with ground truth BRDF parameters and relighting results. Our experiments demonstrate that our algorithm achieves state-of-the-art performance in inverse rendering and relighting, with particularly strong results in the reconstruction of highly reflective objects., Comment: https://dehezhang2.github.io/RISE-SDF/

The current paradigm for the co-evolution of galaxies and their supermassive black holes postulates that dust-obscured active galactic nuclei (AGNs) represent a transitional phase towards a more luminous and unobscured state. However, our understanding of dusty AGNs and their host galaxies at early cosmic times is inadequate due to observational limitations. Here, we present JWST observations of CID-931, an X-ray-detected AGN at a spectroscopic redshift of $z_{\rm spec}=4.91$. Multiband NIRCam imaging from the COSMOS-Web program reveals an unresolved red core, similar to JWST-discovered dusty AGNs. Strikingly, the red core is surrounded by at least eight massive star-forming clumps spread over $1.\!\!^{\prime\prime}6 \approx 10~{\rm kpc}$, each of which has a stellar mass of $10^9-10^{10}M_\odot$ and $\sim0.1-1~{\rm kpc}$ in radius. The whole system amounts to $10^{11}M_\odot$ in stellar mass, higher than typical star-forming galaxies at the same epoch. In this system, gas inflows and/or complex merger events may trigger clump formation and AGN activity thus leading to the rapid formation of a massive galaxy hosting a supermassive black hole. Future follow-up observations will provide new insights into the evolution of the galaxy-black hole relationship during such transitional phases in the early universe., Comment: 13 pages, 10 figures, submitted to PASJ

We study the impact of a non-uniform magnetic background field on the Chiral Magnetic Effect (CME) in equilibrium QCD using lattice simulations with 2+1 flavors of dynamical staggered quarks at the physical point. We show that in the presence of a non-uniform magnetic field the CME manifests itself via a localized electromagnetic current density along the direction of the field, which integrates to zero over the full volume. Our primary observable is the leading-order coefficient of the current in a chiral chemical potential expansion, which we compute for various lattice spacings and extrapolate to the continuum limit. Our findings demonstrate that, even though the global spatial average of the CME conductivity vanishes in equilibrium, steady currents still exist locally. Thus, spatially modulated magnetic fields provide a possible way of generating a non-trivial CME signal in equilibrium., Comment: 10 pages, 8 figures

We propose a novel method for interpreting neural networks, focusing on convolutional neural network-based receiver model. The method identifies which unit or units of the model contain most (or least) information about the channel parameter(s) of the interest, providing insights at both global and local levels -- with global explanations aggregating local ones. Experiments on link-level simulations demonstrate the method's effectiveness in identifying units that contribute most (and least) to signal-to-noise ratio processing. Although we focus on a radio receiver model, the method generalizes to other neural network architectures and applications, offering robust estimation even in high-dimensional settings., Comment: 7+1 pages, 8 figures

Polynomials of degree $\frac{p-1}{2}$ of range sum $p$ were determined by the second authors relying on a joint work of the authors by Kiss and Nagy. We prove that for large enough primes there is no polynomial of degree $\frac{p+1}{2}$ of range sum $p$.

We derive a generalized expressions for the scalar power spectrum and the observational parameters of inflation (the scalar spectral index $n_{\rm s}$ and the tensor-to-scalar ratio $r$) for several single field scalar models minimally coupled with gravity. We work in the framework where the perturbed Einstein equations, at linear order, are of the same form to those obtained in the standard \textit{k}-inflation, with generalized form of the pump field $z$. By generalizing the expressions for $z$ and the speed of sound squared $c_{\rm s}^2$, using the slow-roll approximation, we calculate $n_{\rm s}$ and $r$ up to the second order in the Hubble flow parameters.

This study extends the examination of the Efficient-Market Hypothesis in Bitcoin market during a five year fluctuation period, from September 1 2017 to September 1 2022, by analyzing 28,739,514 qualified tweets containing the targeted topic "Bitcoin". Unlike previous studies, we extracted fundamental keywords as an informative proxy for carrying out the study of the EMH in the Bitcoin market rather than focusing on sentiment analysis, information volume, or price data. We tested market efficiency in hourly, 4-hourly, and daily time periods to understand the speed and accuracy of market reactions towards the information within different thresholds. A sequence of machine learning methods and textual analyses were used, including measurements of distances of semantic vector spaces of information, keywords extraction and encoding model, and Light Gradient Boosting Machine (LGBM) classifiers. Our results suggest that 78.06% (83.08%), 84.63% (87.77%), and 94.03% (94.60%) of hourly, 4-hourly, and daily bullish (bearish) market movements can be attributed to public information within organic tweets., Comment: 33 pages, Preprint

Foundations of formal languages, as subfield of theoretical computer science, are part of typical upper secondary education curricula. There is very little research on the potential difficulties that students at this level have with this subject. In this paper, we report on an exploratory study of errors in formal languages among upper secondary education students. We collect the data by posing exercises in an intelligent tutoring system and analyzing student input. Our results suggest a) instances of non-functional understanding of concepts such as the empty word or a grammar as a substitution system; b) strategic problems such as lack of foresight when deriving a word or confounding formal specifications with real-world knowledge on certain aspects; and c) various syntactic problems. These findings can serve as a starting point for a broader understanding of how and why students struggle with this topic.

For the generalized eigenvalue problem, a quotient function is devised for estimating eigenvalues in terms of an approximate eigenvector. This gives rise to an infinite family of quotients, all entirely arguable to be used in estimation. Although the Rayleigh quotient is among them, one can suggest using it only in an auxiliary manner for choosing the quotient for near optimal results. In normal eigenvalue problems, for any approximate eigenvector, there always exists a "perfect" quotient exactly giving an eigenvalue. For practical estimates in the self-adjoint case, an approximate midpoint of the spectrum is a good choice for reformulating the eigenvalue problem yielding apparently the fastest quotient iterative method there exists. No distinction is made between estimating extreme or interior eigenvalues. Preconditioning from the left results in changing the inner-product and affects the estimates accordingly. Preconditioning from the right preserves self-adjointness and can hence be performed without any restrictions. It is used in variational methods for optimally computing approximate eigenvectors.

Motivated by the burning and cooling processes, the burning game is introduced. The game is played on a graph $G$ by the two players (Burner and Staller) that take turns selecting vertices of $G$ to burn; as in the burning process, burning vertices spread fire to unburned neighbors. Burner aims to burn all vertices of $G$ as quickly as possible, while Staller wants the process to last as long as possible. If both players play optimally, then the number of time steps needed to burn the whole graph $G$ is the game burning number $b_g(G)$ if Burner makes the first move, and the Staller-start game burning number $b_g'(G)$ if Staller starts. In this paper, basic bounds on $b_g(G)$ are given and Continuation Principle is established. Graphs with small game burning numbers are characterized and Nordhaus-Gaddum type results are obtained. An analogue of the burning number conjecture for the burning game is considered and graph products are studied.

Digitizing 3D static scenes and 4D dynamic events from multi-view images has long been a challenge in computer vision and graphics. Recently, 3D Gaussian Splatting (3DGS) has emerged as a practical and scalable reconstruction method, gaining popularity due to its impressive reconstruction quality, real-time rendering capabilities, and compatibility with widely used visualization tools. However, the method requires a substantial number of input views to achieve high-quality scene reconstruction, introducing a significant practical bottleneck. This challenge is especially severe in capturing dynamic scenes, where deploying an extensive camera array can be prohibitively costly. In this work, we identify the lack of spatial autocorrelation of splat features as one of the factors contributing to the suboptimal performance of the 3DGS technique in sparse reconstruction settings. To address the issue, we propose an optimization strategy that effectively regularizes splat features by modeling them as the outputs of a corresponding implicit neural field. This results in a consistent enhancement of reconstruction quality across various scenarios. Our approach effectively handles static and dynamic cases, as demonstrated by extensive testing across different setups and scene complexities., Comment: ECCV 2024 paper. The project page and code are available at https://markomih.github.io/SplatFields/

We analyze JWST MIRI/MRS IFU observations of three Seyferts and showcase the intriguing polycyclic aromatic hydrocarbon (PAH) emission characteristics in regions of $\sim 500\,\rm pc$ scales over or around their active galactic nuclei (AGN). Combining the model predictions and the measurements of PAH features and other infrared emission lines, we find that the central regions containing a high fraction of neutral PAHs with small sizes, e.g., those in ESO137-G034, are in highly heated environments, due to collisional shock heating, with hard and moderately intense radiation fields. Such environments are proposed to be associated with inhibited growth or preferential erosion of PAHs, decreasing the average PAH size and the overall abundance of PAHs. We additionally find that the central regions containing a high fraction of ionized PAHs with large sizes, e.g., those in MCG-05-23-016, are likely experiencing severe photo-ionization because of the radiative effects from the radiative shock precursor besides the AGN. The severe photo-ionization can contribute to the ionization of all PAHs and further destruction of small PAHs. Overall, different Seyferts, even different regions in the same galaxy, e.g., those in NGC\,3081, can contain PAH populations of different properties. Specifically, Seyferts that exhibit similar PAH characteristics to ESO137-G034 and MCG-05-23-016 also tend to have similar emission line properties to them, suggesting that the explanations for PAH characteristics of ESO137-G034 and MCG-05-23-016 may also apply generally. These results have promising application in the era of JWST, especially in diagnosing different (i.e., radiative, and kinetic) AGN feedback modes., Comment: ApJL accepted on September 26th, title slightly modified in accordance with ApJL standards, Fig. 2 updated with additional labels

Utilizing JWST MIRI/MRS IFU observations of the kiloparsec scale central regions, we showcase the diversity of ionized gas distributions and kinematics in six nearby Seyfert galaxies included in the GATOS survey. Specifically, we present spatially resolved flux distribution and velocity field maps of six ionized emission lines covering a large range of ionization potentials ($15.8-97.1$ eV). Based on these maps, we showcase the evidence of ionized gas outflows in the six targets, and find some highly disturbed regions in NGC\,5728, NGC\,5506, and ESO137-G034. We propose AGN-driven radio jets plausibly play an important role in triggering these highly disturbed regions. With the outflow rates estimated based on [Ne~{\footnotesize V}] emission, we find the six targets tend to have ionized outflow rates converged to a narrower range than previous finding. These results have important implication for the outflow properties in AGN of comparable luminosity., Comment: 34 pages (11 pages in the appendix), 18 figures in the main text, ApJ in press (accepted on July 26th)

Fluids at supercritical pressures exhibit large variations in density near the pseudo critical line, such that buoyancy plays a crucial role in their fluid dynamics. Here, we experimentally investigate heat transfer and turbulence in horizontal hydrodynamically developed channel flows of carbon dioxide at 88.5 bar and 32.6{\deg}C, heated at either the top or bottom surface to induce a strong vertical density gradient. In order to visualise the flow and evaluate its heat transfer, shadowgraphy is used concurrently with surface temperature measurements. With moderate heating, the flow is found to strongly stratify for both heating configurations, with bulk Richardson numbers Ri reaching up to 100. When the carbon dioxide is heated from the bottom upwards, the resulting unstably stratified flow is found to be dominated by the increasingly prevalent secondary motion of thermal plumes, enhancing vertical mixing and progressively improving heat transfer compared to a neutrally buoyant setting. Conversely, stable stratification, induced by heating from the top, suppresses the vertical motion leading to deteriorated heat transfer that becomes invariant to the Reynolds number. The optical results provide novel insights into the complex dynamics of the directionally dependent heat transfer in the near-pseudo-critical region. These insights contribute to the reliable design of heat exchangers with highly property-variant fluids, which are critical for the decarbonisation of power and industrial heat. However, the results also highlight the need for further progress in the development of experimental techniques to generate reliable reference data for a broader range of non-ideal supercritical conditions.

Although the importance of the quantum-classical correspondence has been recognized in numerous studies of quantum chaos, whether it still holds for finite time dynamics remains less known. We address this question in this work by performing a detailed analysis of how the quantum chaotic measure relates to the chaoticity of the finite time classical trajectories. A good correspondence between them has been revealed in both time dependent and many-body systems. In particular, we show that the dependence of the quantum chaotic measure on the chaoticity of finite time trajectories can be well captured by a function that is independent of the system. This strongly implies the universal validity of the finite time quantum-classical correspondence. Our findings provide a deeper understanding of the quantum-classical correspondence and highlight the role of time for studying quantum ergodicity., Comment: 13pages, 6 figures

We consider the overfitting behavior of minimum norm interpolating solutions of Gaussian kernel ridge regression (i.e. kernel ridgeless regression), when the bandwidth or input dimension varies with the sample size. For fixed dimensions, we show that even with varying or tuned bandwidth, the ridgeless solution is never consistent and, at least with large enough noise, always worse than the null predictor. For increasing dimension, we give a generic characterization of the overfitting behavior for any scaling of the dimension with sample size. We use this to provide the first example of benign overfitting using the Gaussian kernel with sub-polynomial scaling dimension. All our results are under the Gaussian universality ansatz and the (non-rigorous) risk predictions in terms of the kernel eigenstructure.

The polynomials of degree $\frac{p-1}{2}$ of range sum $p$ was determined in {\tt arXiv:2311.06136 [math.NT]} for large enough primes. We extend this result by reducing the lower bound for the primes to $23$ by introducing a new and elementary way of estimating sums of Legendre symbols., Comment: 12 pages

Cold gas and dust reprocess the X-ray emission of active galactic nuclei (AGN), producing spectro-polarimetric features in the X-ray band. The recent launch of IXPE allows for observations of this X-ray polarisation signal, encoding unique information on the circumnuclear medium of AGN. However, the models for interpreting these polarimetric data are under-explored and do not reach the same level of sophistication as the corresponding spectral models. We aim at closing the gap between the spectral and spectro-polarimetric modelling of AGN circumnuclear media by providing the tools for simulating X-ray polarisation in complex 3D transfer media alongside X-ray spectra. We lay out the framework for X-ray polarisation in 3D radiative transfer simulations and provide an implementation to the 3D radiative transfer code SKIRT, focussing on (de)polarisation due to scattering and fluorescent re-emission. As a first application, we studied a 2D toroidal reprocessor of cold gas, modelling the AGN circumnuclear medium. For the 2D torus model, we find a complex behaviour of the polarisation angle with photon energy, which we interpret as a balance between the reprocessed photon flux originating from different sky regions, with a direct link to the torus geometry. We calculated a large grid of AGN torus models and demonstrated how spatially resolved polarisation maps could form a useful tool for interpreting the geometrical information that is encoded in IXPE observations. With this work, we release high-resolution AGN torus templates that simultaneously describe X-ray spectra and spectro-polarimetry, for observational data fitting with XSPEC. The SKIRT code can now model X-ray polarisation simultaneously with X-ray spectra and provide synthetic spectro-polarimetric observations for complex 3D circumnuclear media, with all features of the established SKIRT framework available., Comment: 16 pages, 13 figures, accepted for publication in Astronomy & Astrophysics

We study spin-orbit proximity effects in an armchair (4,4) carbon nanotube on the Pt(111) surface. By employing first-principles calculations, we show that the Dirac cone of the metallic nanotube is altered due to strong hybridization with the Pt substrate. Inserting a monolayer hexagonal boron nitride (hBN) between the nanotube and the substrate limits the hybridization effects leading to recovering the Dirac cone. The Dirac bands display asymmetric spin splitting, 0.7\,meV for the right movers and 1.7\,meV for the left movers at the K valley, due to the proximity to the Pt substrate. We find that the Dirac states exhibit almost perfect spin polarization, transverse to the nanotube axis and to the stacking direction, forming a proper condition for charge-to-spin conversion with coherent spin transport in the nanotube. We propose an effective Hamiltonian describing the proximity-induced effects on the Dirac electrons and their spin texture.

We present the first determination of the topological susceptibility from lattice QCD in the presence of strong background magnetic fields. Our simulations employ 2+1 flavours of stout improved staggered quarks with physical masses and cover a broad range of temperatures and magnetic field values. The results are extrapolated to the continuum limit using four different lattice spacings and an eigenvalue reweighting technique to reduce discretisation errors. For low temperatures, our calculations show an enhancement of the topological susceptibility due to the magnetic field, compatible with predictions from chiral perturbation theory. At high temperatures, we observe the impact of inverse magnetic catalysis on the susceptibility., Comment: 20 pages, 10 figures

The use of large language models (LLMs) is becoming increasingly widespread among software developers. However, privacy and computational requirements are problematic with commercial solutions and the use of LLMs. In this work, we focus on using LLMs with around 160 million parameters that are suitable for local execution and augmentation with retrieval from local projects. We train two models based on the transformer architecture, the generative model GPT-2 and the retrieval-adapted RETRO model, on open-source Python files, and empirically evaluate and compare them, confirming the benefits of vector embedding based retrieval. Further, we improve our models' performance with In-context retrieval-augmented generation, which retrieves code snippets based on the Jaccard similarity of tokens. We evaluate In-context retrieval-augmented generation on larger models and conclude that, despite its simplicity, the approach is more suitable than using the RETRO architecture. We highlight the key role of proper tokenization in achieving the full potential of LLMs in code completion., Comment: 28 pages, 14 figures

Quantum transport studies of spin-dependent phenomena in solids commonly employ the Kubo or Keldysh formulas for the steady-state density matrix in the linear-response regime. Its trace with operators of interest -- such as, spin density, spin current density or spin torque -- gives expectation values of experimentally accessible observables. For such local quantities, these formulas require summing over the manifolds of {\em both} Fermi-surface and Fermi-sea quantum states. However, debates have been raging in the literature about vastly different physics the two formulations can apparently produce, even when applied to the same system. Here, we revisit this problem using a testbed of infinite-size graphene with proximity-induced spin-orbit and magnetic exchange effects. By splitting this system into semi-infinite leads and central active region, in the spirit of Landauer two-terminal setup for quantum transport, we prove the {\em numerically exact equivalence} of the Kubo and Keldysh approaches via the computation of spin Hall current density and spin-orbit torque in both clean and disordered limits. The key to reconciling the two approaches are the numerical frameworks we develop for: ({\em i}) evaluation of Kubo(-Bastin) formula for a system attached to semi-infinite leads, which ensure continuous energy spectrum and evade the need for phenomenological broadening in prior calculations; and ({\em ii}) proper evaluation of Fermi-sea term in the Keldysh approach, which {\em must} include the voltage drop across the central active region even if it is disorder free., Comment: 9 pages, 3 figures, 111 references

This paper argues that military buildups lead to a significant rise in greenhouse gas emissions and can disrupt the green transition. Identifying military spending shocks, I use local projections to show that a percentage point rise in the military spending share leads to a 0.9-2% rise in total emissions, as well as a 1% rise in emission intensity, depending on the economy's overall emission intensity. Using a dynamic production network model calibrated for the US, I find that a permanent shock of the same size would increase total emissions by between 0.36% and 1.81%, and emission intensity by between 0.22% and 1.5%. The empirical analysis indicates that green patenting is reduced by 10-25% following such a shock, and the model suggests that investment in renewables could be crowded out by defence spending, hindering the energy transition. These effects can significantly raise climate damages, temperatures, and can lead to a quicker exhaustion of the carbon budget.

Advancements in artificial intelligence have driven the development of numerous pathology foundation models capable of extracting clinically relevant information. However, there is currently limited literature independently evaluating these foundation models on truly external cohorts and clinically-relevant tasks to uncover adjustments for future improvements. In this study, we benchmarked ten histopathology foundation models on 13 patient cohorts with 6,791 patients and 9,493 slides from lung, colorectal, gastric, and breast cancers. The models were evaluated on weakly-supervised tasks related to biomarkers, morphological properties, and prognostic outcomes. We show that a vision-language foundation model, CONCH, yielded the highest performance in 42% of tasks when compared to vision-only foundation models. The experiments reveal that foundation models trained on distinct cohorts learn complementary features to predict the same label, and can be fused to outperform the current state of the art. Creating an ensemble of complementary foundation models outperformed CONCH in 66% of tasks. Moreover, our findings suggest that data diversity outweighs data volume for foundation models. Our work highlights actionable adjustments to improve pathology foundation models.