Search

In this manuscript we define the notion of "$\delta$-typicality" for both entropy and relative entropy, as well as a notion of $\epsilon$-goodness and provide an extension to Stein's lemma for continuous quantities as well as correlated setups. We apply the derived results on the Gaussian hypothesis testing problem where the observations are possibly correlated., Comment: 11 pages

The fall 2023 and spring 2024 undergraduate enrollment increases, marking the first growth since the COVID-19 pandemic, show signs of a post-pandemic turnaround for higher education. However, a significant share of current undergraduates will eventually disengage from college before earning a degree or other credential. They will join tens of millions of other adult Americans who are Some College, No Credential (SCNC). The SCNC population has been consistently rising over time. Re-engaging those who stop out remains a persistent challenge and a priority for the forty states that have set ambitious postsecondary attainment goals. This report aims to provide timely insights into the SCNC population, offering state leaders and policymakers accurate data on its current status, along with tracking progress and outcome measures for SCNC students. The first section of this report describes who makes up the SCNC population and how it has changed since the last report. In this section, the authors pay particular attention to Recent Stopouts, who joined the SCNC population after being stopped out between January 2021 and July 2022. In the second section, the authors report on SCNC re-enrollment in the 2022-23 academic year as well as first-year credential earning for re-enrollees. The authors also provide new updates on continued enrollment and second-year credential earning for SCNC re-enrollees in the 2021-22 academic year, whom were first reported on last year.

Concept-based explanation methods, such as concept bottleneck models (CBMs), aim to improve the interpretability of machine learning models by linking their decisions to human-understandable concepts, under the critical assumption that such concepts can be accurately attributed to the network's feature space. However, this foundational assumption has not been rigorously validated, mainly because the field lacks standardised metrics and benchmarks to assess the existence and spatial alignment of such concepts. To address this, we propose three metrics: the concept global importance metric, the concept existence metric, and the concept location metric, including a technique for visualising concept activations, i.e., concept activation mapping. We benchmark post-hoc CBMs to illustrate their capabilities and challenges. Through qualitative and quantitative experiments, we demonstrate that, in many cases, even the most important concepts determined by post-hoc CBMs are not present in input images; moreover, when they are present, their saliency maps fail to align with the expected regions by either activating across an entire object or misidentifying relevant concept-specific regions. We analyse the root causes of these limitations, such as the natural correlation of concepts. Our findings underscore the need for more careful application of concept-based explanation techniques especially in settings where spatial interpretability is critical., Comment: 17 pages it total, 8 main pages

Over the last decade, genome-wide association studies (GWAS) have successfully identified numerous genetic variants associated with complex diseases. These associations have the potential to reveal the molecular mechanisms underlying complex diseases and lead to the identification of novel drug targets. Despite these advancements, the biological pathways and mechanisms linking genetic variants to complex diseases are still not fully understood. Most trait-associated variants reside in non-coding regions and are presumed to influence phenotypes through regulatory effects on gene expression. Yet, it is often unclear which genes they regulate and in which cell types this regulation occurs. Transcriptome-wide association studies (TWAS) aim to bridge this gap by detecting trait-associated tissue gene expression regulated by GWAS variants. However, traditional TWAS approaches frequently overlook the critical contributions of trans-regulatory effects and fail to integrate comprehensive regulatory networks. Here, we present a novel framework that leverages tissue-specific gene regulatory networks (GRNs) to integrate cis- and trans-genetic regulatory effects into the TWAS framework for complex diseases. We validate our approach using coronary artery disease (CAD), utilizing data from the STARNET project, which provides multi-tissue gene expression and genetic data from around 600 living patients with cardiovascular disease. Preliminary results demonstrate the potential of our GRN-driven framework to uncover more genes and pathways that may underlie CAD. This framework extends traditional TWAS methodologies by utilizing tissue-specific regulatory insights and advancing the understanding of complex disease genetic architecture., Comment: 22 pages, 6 figures, 4 tables; code available at https://github.com/guutama/GRN-TWAS

Open Radio Access Network (O-RAN) along with artificial intelligence, machine learning, cloud and edge networking, and virtualization are important enablers for designing flexible and software-driven programmable wireless networks. In addition, Reconfigurable Intelligent Surfaces (RIS) represent an innovative technology to direct incoming radio signals toward desired locations by software-controlled passive reflecting antenna elements. Despite their distinctive potential, there has been limited exploration of integrating RIS with the O-RAN framework, an area that holds promise for enhancing next-generation wireless systems. This paper addresses this gap by designing and developing the RIS optimization xApps within an O-RAN-based real-time 5G environment. We perform extensive measurement experiments using an end-to-end 5G testbed including the RIS prototype in a multi-user scenario. The results demonstrate that the RIS can be utilized either to boost the performance of the selected user or to provide the fairness among the users or to balance the tradeoff between the performance and fairness., Comment: 6 pages, 7 figures

In this paper, we derive a periodic model from a one dimensional nonlocal eikonal equation set on the full space modeling dislocation dynamics. Thanks to a gradient entropy estimate, we show that this periodic model converges toward the initial one when the period goes to infinity. Moreover, we design a semi-explicit numerical scheme for the periodic model that we introduce. We show the well-posedness of the scheme and a discrete gradient entropy inequality. We also prove the convergence of the scheme and we present some numerical experiments.

The increase in video streaming has presented a challenge of handling stream request effectively, especially over networks that are variable. This paper describes a new adaptive video streaming architecture capable of changing the video quality and buffer size depending on the data and latency of streamed video. For video streaming VLC media player was used where network performance data were obtained through Python scripts with very accurate data rate and latency measurement. The collected data is analyzed using Gemini AI, containing characteristics of the machine learning algorithm that recognizes the best resolution of videos and the buffer sizes. Through the features of real-time monitoring and artificial intelligence decision making, the proposed framework improves the user experience by reducing the occurrence of buffering events while at the same time increasing the video quality. Our findings therefore confirm that the proposed solution based on artificial intelligence increases video quality and flexibility. This study advances knowledge of adaptive streaming and offers an argument about how intelligent datadriven approaches and AI may be useful tools for enhancing the delivery of video in practical environments.

Bluetooth Low Energy (BLE) location trackers, or "tags", are popular consumer devices for monitoring personal items. These tags rely on their respective network of companion devices that are capable of detecting their BLE signals and relay location information back to the owner. While manufacturers claim that such crowd-sourced approach yields accurate location tracking, the tags' real-world performance characteristics remain insufficiently understood. To this end, this study presents a comprehensive analysis of three major players in the market: Apple's AirTag, Samsung's SmartTag, and Tile. Our methodology combines controlled experiments -- with a known large distribution of location-reporting devices -- as well as in-the-wild experiments -- with no control on the number and kind of reporting devices encountered, thus emulating real-life use-cases. Leveraging data collection techniques improved from prior research, we recruit 22 volunteers traveling across 29 countries, examining the tags' performance under various environments and conditions. Our findings highlight crucial updates in device behavior since previous studies, with AirTag showing marked improvements in location report frequency. Companion device density emerged as the primary determinant of tag performance, overshadowing technological differences between products. Additionally, we find that post-COVID-19 mobility trends could have contributed to enhanced performance for AirTag and SmartTag. Tile, despite its cross-platform compatibility, exhibited notably lower accuracy, particularly in Asia and Africa, due to limited global adoption. Statistical modeling of spatial errors -- measured as the distance between reported and actual tag locations -- shows log-normal distributions across all tags, highlighting the need for improved location estimation methods to reduce occasional significant inaccuracies., Comment: 13 pages, 16 Figures, 7 Tables

TikTok is a major force among social media platforms with over a billion monthly active users worldwide and 170 million in the United States. The platform's status as a key news source, particularly among younger demographics, raises concerns about its potential influence on politics in the U.S. and globally. Despite these concerns, there is scant research investigating TikTok's recommendation algorithm for political biases. We fill this gap by conducting 323 independent algorithmic audit experiments testing partisan content recommendations in the lead-up to the 2024 U.S. presidential elections. Specifically, we create hundreds of "sock puppet" TikTok accounts in Texas, New York, and Georgia, seeding them with varying partisan content and collecting algorithmic content recommendations for each of them. Collectively, these accounts viewed ~394,000 videos from April 30th to November 11th, 2024, which we label for political and partisan content. Our analysis reveals significant asymmetries in content distribution: Republican-seeded accounts received ~11.8% more party-aligned recommendations compared to their Democratic-seeded counterparts, and Democratic-seeded accounts were exposed to ~7.5% more opposite-party recommendations on average. These asymmetries exist across all three states and persist when accounting for video- and channel-level engagement metrics such as likes, views, shares, comments, and followers, and are driven primarily by negative partisanship content. Our findings provide insights into the inner workings of TikTok's recommendation algorithm during a critical election period, raising fundamental questions about platform neutrality., Comment: 19 pages, 5 Figures, 1 Table

Mixed Reality (MR) devices are being increasingly adopted across a wide range of real-world applications, ranging from education and healthcare to remote work and entertainment. However, the unique immersive features of MR devices, such as 3D spatial interactions and the encapsulation of virtual objects by invisible elements, introduce new vulnerabilities leading to interaction obstruction and misdirection. We implemented latency, click redirection, object occlusion, and spatial occlusion attacks within a remote collaborative MR platform using the Microsoft HoloLens 2 and evaluated user behavior and mitigations through a user study. We compared responses to MR-specific attacks, which exploit the unique characteristics of remote collaborative immersive environments, and traditional security attacks implemented in MR. Our findings indicate that users generally exhibit lower recognition rates for immersive attacks (e.g., spatial occlusion) compared to attacks inspired by traditional ones (e.g., click redirection). Our results demonstrate a clear gap in user awareness and responses when collaborating remotely in MR environments. Our findings emphasize the importance of training users to recognize potential threats and enhanced security measures to maintain trust in remote collaborative MR systems., Comment: 11 pages, 5 figures, 2 tables, in proceedings of IEEE VR 2025

We investigate a Kyle model under Gaussian assumptions where a risk-averse informed trader has imperfect information on the fundamental price of an asset. We show that an equilibrium can be constructed by considering an optimal transport problem that is solved under a measure that renders the utility of the informed trader martingale and a filtering problem under the historical measure.

This study addresses a critical gap in the healthcare system by developing a clinically meaningful, practical, and explainable disease surveillance system for multiple chronic diseases, utilizing routine EHR data from multiple U.S. practices integrated with CureMD's EMR/EHR system. Unlike traditional systems--using AI models that rely on features from patients' labs--our approach focuses on routinely available data, such as medical history, vitals, diagnoses, and medications, to preemptively assess the risks of chronic diseases in the next year. We trained three distinct models for each chronic disease: prediction models that forecast the risk of a disease 3, 6, and 12 months before a potential diagnosis. We developed Random Forest models, which were internally validated using F1 scores and AUROC as performance metrics and further evaluated by a panel of expert physicians for clinical relevance based on inferences grounded in medical knowledge. Additionally, we discuss our implementation of integrating these models into a practical EMR system. Beyond using Shapley attributes and surrogate models for explainability, we also introduce a new rule-engineering framework to enhance the intrinsic explainability of Random Forests.

The significance of the automatic ground collision avoidance system (Auto-GCAS) has been proven by considering the fatal crashes that have occurred over decades. Even though extensive efforts have been put forth to address the ground collision avoidance in the literature, the notion of being nuisance-free has not been sufficiently addressed. At this point, in this study, the Auto-GCAS design is formulated by merging exponential control barrier functions with sliding manifolds to manipulate the barrier function dynamics. The adaptive properties of the sliding manifolds are tailored to the key and governing flight parameters, ensuring that the nuisance-free requirement is satisfied. Furthermore, to ensure all safety requirements are met, a flight envelope protection algorithm is designed using control barrier functions to assess the commands generated by the Auto-GCAS. Eventually, the performance of the proposed methodology is demonstrated, focusing on authority-sharing, collision avoidance capability, and nuisance-free operation through various scenarios and Monte Carlo simulations.

In recent years, many attempts have been made to enhance Orthogonal Frequency Multiplexing with Index Modulation (OFDM-IM) in terms of spectral efficiency and error performance. Two challenges typically erupt when using OFDM-IM. First, the degradation in spectral efficiency due to the subcarrier's deactivation, especially when using higher order modulation (M-ary) where every inactive subcarrier will cost $Log_2(M)$ bits loss. Second, using a fixed number of active subcarriers within a sub-block forces the error to be localized within the sub-block. Yet, it loses the advantage of exploiting all possible pattern combinations degrading the overall spectral efficiency. In this paper, we introduce a solution to tackle those problems. The Enhanced Generalized Index Modulation (EGIM) is a simple systematic way to generate and detect the OFDM-IM frame. Unlike the classical OFDM-IM generation by splitting the frame into sub-frames which increases the complexity of the OFDM-IM transmitter and reflects on the receiver Maximum likelihood detector, EGIM Makes full use of all possible combinations of active subcarriers within the frame by using variable active subcarriers (k) depending on the incoming data. The EGIM is still susceptible to error propagation if the OFF symbol is wrongly mapped to one of the ON symbols or vice versa. For that reason, we offer an OFDM-IM autoencoder to overcome this problem. The encoder generates the (ON/OFF) symbols systematically to achieve the advantage of sending all possible frame indices patterns depending on the input bit stream offering an average of 3dB gain in terms of power efficiency. The proposed encoder performance was compared to the standard encoder with the same effective coding rate using soft and hard decision Viterbi decoding utilizing the power gain achieved., Comment: 10 Pages, 12 Figures

This paper presents a novel approach for enhancing wireless signal reception through self-adjustable metallic surfaces, termed reflectors, which are guided by deep reinforcement learning (DRL). The designed reflector system aims to improve signal quality for multiple users in scenarios where a direct line-of-sight (LOS) from the access point (AP) and reflector to users is not guaranteed. Utilizing DRL techniques, the reflector autonomously modifies its configuration to optimize beam allocation from the AP to user equipment (UE), thereby maximizing path gain. Simulation results indicate substantial improvements in the average path gain for all UEs compared to baseline configurations, highlighting the potential of DRL-driven reflectors in creating adaptive communication environments.

Accurately predicting electronic band gaps in halide perovskites using ab initio density functional theory (DFT) is essential for their application in optoelectronic devices. Standard hybrid functionals such as HSE and PBE0 can overcome the limitations of DFT with reasonable computational cost but are known to underestimate the measured band gaps for layered halide perovskites. In this study, we assess the performance of the doubly screened dielectric-dependent hybrid (DSH) functional for predicting band gaps in three-dimensional (3D) and layered hybrid perovskites. We show that the DSH functional, which employs material-dependent mixing parameters derived from macroscopic dielectric constants, provides accurate band gap predictions for 3D halide perovskites when structural local disorder is considered. For layered hybrid perovskites, DSH functional based on average dielectric constants overestimates the band gaps. To improve the predictions and stay in a parameter-free ab initio workflow, we propose to use the calculated dielectric constant of the respective 3D perovskites. We find that the DSH functionals using the dielectric constants of the 3D perovskite accurately predict experimental gaps, with the lowest mean absolute errors compared to HSE and PBE0 for layered perovskites with various organic spacers, as well as for multilayered $BA_2MA_{n-1}Pb_{n}I_{3n-1}$ with n = 2, 3. Notably, the HSE functional systematically underestimates the band gaps in layered perovskites. We attribute the root of this failure to the absence of non-local long-range dielectric screening, a critical factor for halide perovskites. The computational framework introduced here provides an efficient parameter-free ab initio methodology for predicting the electronic properties of 3D and layered halide perovskites and their heterostructures, aiding in developing advanced optoelectronic devices.

Reconfigurable intelligent surfaces (RISs) are eminently suitable for improving the reliability of wireless communications by jointly designing the active beamforming at the base station (BS) and the passive beamforming at the RIS. Therefore, the accuracy of channel estimation is crucial for RIS-aided systems. The challenge is that only the cascaded two-hop channel spanning from the user equipments (UEs) to the RIS and spanning from the RIS to the BS can be estimated, due to the lack of active radio frequency (RF) chains at RIS elements, which leads to high pilot overhead. In this paper, we propose a low-overhead linear minimum mean square error (LMMSE) channel estimation method by exploiting the spatial correlation of channel links, which strikes a trade-off between the pilot overhead and the channel estimation accuracy. Moreover, we calculate the theoretical normalized mean square error (MSE) for our channel estimation method. Finally, we verify numerically that the proposed LMMSE estimator has lower MSE than the state-of-the-art (SoA) grouping based estimators.

The proliferation of web agents necessitates advanced navigation and interaction strategies within complex web environments. Current models often struggle with efficient navigation and action execution due to limited visibility and understanding of web structures. Our proposed R2D2 framework addresses these challenges by integrating two paradigms: Remember and Reflect. The Remember paradigm utilizes a replay buffer that aids agents in reconstructing the web environment dynamically, thus enabling the formulation of a detailed ``map'' of previously visited pages. This helps in reducing navigational errors and optimizing the decision-making process during web interactions. Conversely, the Reflect paradigm allows agents to learn from past mistakes by providing a mechanism for error analysis and strategy refinement, enhancing overall task performance. We evaluate R2D2 using the WEBARENA benchmark, demonstrating significant improvements over existing methods, including a 50% reduction in navigation errors and a threefold increase in task completion rates. Our findings suggest that a combination of memory-enhanced navigation and reflective learning promisingly advances the capabilities of web agents, potentially benefiting various applications such as automated customer service and personal digital assistants.

Novel capacities of generative AI to analyze and generate cultural artifacts raise inevitable questions about the nature and value of artistic education and human expertise. Has AI already leveled the playing field between professional artists and laypeople, or do trained artistic expressive capacity, curation skills and experience instead enhance the ability to use these new tools? In this pre-registered study, we conduct experimental comparisons between 50 active artists and a demographically matched sample of laypeople. We designed two tasks to approximate artistic practice for testing their capabilities in both faithful and creative image creation: replicating a reference image, and moving as far away as possible from it. We developed a bespoke platform where participants used a modern text-to-image model to complete both tasks. We also collected and compared participants' sentiments towards AI. On average, artists produced more faithful and creative outputs than their lay counterparts, although only by a small margin. While AI may ease content creation, professional expertise is still valuable - even within the confined space of generative AI itself. Finally, we also explored how well an exemplary vision-capable large language model (GPT-4o) would complete the same tasks, if given the role of an image generation agent, and found it performed on par in copying but outperformed even artists in the creative task. The very best results were still produced by humans in both tasks. These outcomes highlight the importance of integrating artistic skills with AI training to prepare artists and other visual professionals for a technologically evolving landscape. We see a potential in collaborative synergy with generative AI, which could reshape creative industries and education in the arts., Comment: Eisenmann and Karjus contributed equally to this work and share first authorship

In ultrathin ferromagnetic films sandwiched between two distinct heavy metal layers or between a heavy metal and an oxide layer, the Dzyaloshinskii-Moriya interaction (DMI) is recognized as being of interfacial origin. Its chirality and strength are determined by the properties of the adjacent heavy metals and the degree of oxidation at the interfaces. Here, we demonstrate that the chirality of the DMI can change solely with variations in the thickness of the ferromagnetic layer - an effect that has not been experimentally observed or explained until now. Our experimental observation in the trilayer system Ta/FeCoB/TaOx is supported by ab initio calculations: they reveal that variations in orbital filling and inter-atomic distances at the interface, driven by the number of ferromagnetic atomic layers, lead to an inversion of DMI chirality. This mechanism takes place for ferromagnetic layers with more than three atomic layers, for which the two interfaces start to be decoupled. We hence propose a new degree of freedom to tune DMI chirality and the associated chiral spin textures by tailoring crystal structure e.g. using strain or surface acoustic waves., Comment: 5 figures + supplementary materials

This study introduces a Block Phase Tracking Reference Signal (PTRS) allocation approach for Discrete Fourier Transform-spread-Orthogonal Frequency Division Multiplexing (DFT-s-OFDM) systems to enhance phase noise tracking and compensation. Our proposed block allocation methodology leverages the concepts of multiresolution time-frequency tiling for more effective sampling, thereby mitigating aliasing effects and improving phase noise resilience. A key contribution of our approach is a novel modulation and demodulation scheme, incorporating a dedicated DFT-s-OFDM symbol, a modulator branch for block PTRS generation, and a dedicated demodulator for accurate phase noise estimation and correction.

Language serves as a foundation of cultural identity, deeply entangled with the social and historical contexts of a community. This paper examines the ineffectiveness of interference by alien words within a culture. Drawing on sociolinguistic theories and case studies from diverse linguistic environments, it is argued that the forced introduction or adoption of foreign lexicon often fails to achieve its intended socio-cultural objectives. Instead, indigenous languages demonstrate resilience, adapting to or resisting external influences through unique strategies. The effectiveness of this research highlights the futility of attempting to impose linguistic uniformity and underscores the importance of understanding local cultural dynamics in preserving linguistic heritage. This pure language understanding directly relates to translation knowledge where linguists and translators need to work and research to eradicate misunderstanding. Misunderstandings mostly appear in non-equivalent words because there are different local and internal words like food, garment, cultural and traditional words, and others in every notion. Truly, most of these words do not have an equivalent in the target language and these words need to be worked and find their equivalent in the target language to fully understand both languages. The purpose of this research is to introduce the challenges and ineffectiveness of cultural influences in different notions where people do not see the facts of cultural enrichment. However, some of these ineffectiveness have been clearly mentioned in this research but some effective ways have also been dictated., Comment: Internal Journal Publication

Tracing the cosmic path of galaxies requires an understanding of their chemical enrichment and merging histories. One of the most important constraints is the internal structure of galaxies, notably the internal distribution of elements acting as fossils in extra-galactic archaeology. Using our cosmological chemodynamical simulations, which include all relevant physical processes and the latest nucleosynthesis yields, we investigate the evolution of radial metallicity gradients of stellar populations and the interstellar medium within each galaxy. This work explores the role of supernova feedback on the metallicity gradients by comparing three feedback models, ejecting energy in thermal, stochastic and mechanical forms. At $z=0$, the mechanical feedback model produces the gradient--mass relations of stars and gas both in excellent agreement with observations; gradients are the steepest at intermediate-mass ($M_*\sim10^{10}M_\odot$) and become flatter in massive galaxies probably by major mergers. For each model, we predict similar gradient--mass relations up to $z=4$ and find that the mechanical feedback model gives flatter gradients of both stars and gas for lower-mass galaxies ($M_*<10^{10}M_\odot$) possibly due to the suppression of star formation and metal ejection by stellar feedback. With all feedback models, most galaxies have negative gas-phase metallicity gradients up to $z=5$, suggesting an inside-out growth, which is consistent with other cosmological simulations but not with recent observations at $z\sim1$--2.5. We find a mild redshift evolution of gradients up to $z=4$, while there seems to be an evolutionary transition at $z=5$ where the metallicity gradients become steep for gas and stars., Comment: 17 pages, 17 figures. Submitted to MNRAS

Objective: Ficus exasperata Vahl (Sandpaper tree) is extensively used in Nigeria to treat diseases, but a dearth of documentation about its toxicity exists. This information is crucial because pollutants can contaminate medicinal plants. This study determined the heavy metal and phytochemical content of methanolic leaf extract of F exasperata obtained in Birnin Kebbi, Nigeria. Material and Methods: The lethality of the plant was also assessed using 70 wild shrimps divided equally into seven groups. Group 1 (negative control), groups 2 and 3 (positive controls) were exposed to 500 and 1000ppm of formaldehyde, respectively; and groups 4-7 were exposed to 1000, 2000, 4000, and 8000ppm of extracts, respectively, for 96 hours. Results: The phytochemistry revealed high levels of flavonoids and saponins and moderate levels of tannins and phenols. The heavy metal analysis revealed non-tolerable levels of cadmium, copper, and lead, while zinc was within the tolerable limit. The negative control recorded 10% mortality, 1000 and 2000 ppm (20% each), 4000ppm (70%), and 8000 ppm (100%). Conclusion: These results inferred safe doses of the plant's extract in low and medium concentrations but toxic and fatal at high doses over a period of time. Consumers are advised to seek an expert's guidance before using it., Comment: 7 pages

This study presents a comprehensive review of the potential of multimodal deep learning (DL) in medical diagnosis, using COVID-19 as a case example. Motivated by the success of artificial intelligence applications during the COVID-19 pandemic, this research aims to uncover the capabilities of DL in disease screening, prediction, and classification, and to derive insights that enhance the resilience, sustainability, and inclusiveness of science, technology, and innovation systems. Adopting a systematic approach, we investigate the fundamental methodologies, data sources, preprocessing steps, and challenges encountered in various studies and implementations. We explore the architecture of deep learning models, emphasising their data-specific structures and underlying algorithms. Subsequently, we compare different deep learning strategies utilised in COVID-19 analysis, evaluating them based on methodology, data, performance, and prerequisites for future research. By examining diverse data types and diagnostic modalities, this research contributes to scientific understanding and knowledge of the multimodal application of DL and its effectiveness in diagnosis. We have implemented and analysed 11 deep learning models using COVID-19 image, text, and speech (ie, cough) data. Our analysis revealed that the MobileNet model achieved the highest accuracy of 99.97% for COVID-19 image data and 93.73% for speech data (i.e., cough). However, the BiGRU model demonstrated superior performance in COVID-19 text classification with an accuracy of 99.89%. The broader implications of this research suggest potential benefits for other domains and disciplines that could leverage deep learning techniques for image, text, and speech analysis., Comment: 43 pages

Foundation models (FMs) have shown transformative potential in radiology by performing diverse, complex tasks across imaging modalities. Here, we developed CT-FM, a large-scale 3D image-based pre-trained model designed explicitly for various radiological tasks. CT-FM was pre-trained using 148,000 computed tomography (CT) scans from the Imaging Data Commons through label-agnostic contrastive learning. We evaluated CT-FM across four categories of tasks, namely, whole-body and tumor segmentation, head CT triage, medical image retrieval, and semantic understanding, showing superior performance against state-of-the-art models. Beyond quantitative success, CT-FM demonstrated the ability to cluster regions anatomically and identify similar anatomical and structural concepts across scans. Furthermore, it remained robust across test-retest settings and indicated reasonable salient regions attached to its embeddings. This study demonstrates the value of large-scale medical imaging foundation models and by open-sourcing the model weights, code, and data, aims to support more adaptable, reliable, and interpretable AI solutions in radiology., Comment: 6 figures, followed by 9 Extended Data Figures and a Supplementary Information document

Deep neural networks (DNNs) remain challenged by distribution shifts in complex open-world domains like automated driving (AD): Absolute robustness against yet unknown novel objects (semantic shift) or styles like lighting conditions (covariate shift) cannot be guaranteed. Hence, reliable operation-time monitors for identification of out-of-training-data-distribution (OOD) scenarios are imperative. Current approaches for OOD classification are untested for complex domains like AD, are limited in the kinds of shifts they detect, or even require supervision with OOD samples. To prepare for unanticipated shifts, we instead establish a framework around a principled, unsupervised, and model-agnostic method that unifies detection of all kinds of shifts: Find a full model of the training data's feature distribution, to then use its density at new points as in-distribution (ID) score. To implement this, we propose to combine the newly available Vision Foundation Models (VFM) as feature extractors with one of four alternative density modeling techniques. In an extensive benchmark of 4 VFMs against 20 baselines, we show the superior performance of VFM feature encodings compared to shift-specific OOD monitors. Additionally, we find that sophisticated architectures outperform larger latent space dimensionality; and our method identifies samples with higher risk of errors on downstream tasks, despite being model-agnostic. This suggests that VFMs are promising to realize model-agnostic, unsupervised, reliable safety monitors in complex vision tasks.

Hate speech and abusive language are global phenomena that need socio-cultural background knowledge to be understood, identified, and moderated. However, in many regions of the Global South, there have been several documented occurrences of (1) absence of moderation and (2) censorship due to the reliance on keyword spotting out of context. Further, high-profile individuals have frequently been at the center of the moderation process, while large and targeted hate speech campaigns against minorities have been overlooked. These limitations are mainly due to the lack of high-quality data in the local languages and the failure to include local communities in the collection, annotation, and moderation processes. To address this issue, we present AfriHate: a multilingual collection of hate speech and abusive language datasets in 15 African languages. Each instance in AfriHate is annotated by native speakers familiar with the local culture. We report the challenges related to the construction of the datasets and present various classification baseline results with and without using LLMs. The datasets, individual annotations, and hate speech and offensive language lexicons are available on https://github.com/AfriHate/AfriHate

Network slicing is a key enabler for providing a differentiated service support to heterogeneous use cases and applications in 5G and beyond networks through creating multiple logical slices. Resource allocation for satisfying diverse requirements of slices is a highly challenging task under time-varying traffic and wireless channel conditions. This paper presents a deep reinforcement learning (DRL) approach for allocating radio resources to slices, where the objective is to meet the latency requirement of the low-latency slice without jeopardizing the performance of the other slice. The proposed DRL approach is implemented within an open source mobile network emulator, namely OpenAirInterface, to create an O-RAN compliant end-to-end 5G network capable of dynamic resource allocation capabilities. The intelligent resource allocation mechanism operates on the RAN Intelligent Controller (RIC) as an xApp, enabling monitoring and dynamic resource control of the gNB through the E2 interface. The results demonstrate that the latency requirement of the low-latency slice is met under extremely loaded traffic scenarios, where the trained DRL model deployed on the near-RT RIC platform is used to dynamically allocate the radio resources to the slices., Comment: 5 pages, 5 figures, conference

This paper evaluates the performance of reconfigurable intelligent surface (RIS) optimization algorithms, which utilize channel estimation methods, in ray tracing (RT) simulations within urban digital twin environments. Beyond Sionna's native capabilities, we implement and benchmark additional RIS optimization algorithms based on channel estimation, enabling an evaluation of RIS strategies under various deployment conditions. Coverage maps for RIS-assisted communication systems are generated through the integration of Sionna's RT simulations. Moreover, real-world experimentation underscores the necessity of validating algorithms in near-realistic simulation environments, as minor variations in measurement setups can significantly affect performance., Comment: Submitted to possible IEEE conference

Traffic flow prediction remains a cornerstone for intelligent transportation systems ITS, influencing both route optimization and environmental efforts. While Recurrent Neural Networks RNN and traditional Convolutional Neural Networks CNN offer some insights into the spatial temporal dynamics of traffic data, they are often limited when navigating sparse and extended spatial temporal patterns. In response, the CNN-GRUSKIP model emerges as a pioneering approach. Notably, it integrates the GRU-SKIP mechanism, a hybrid model that leverages the Gate Recurrent Unit of GRU capabilities to process sequences with the SKIP feature of ability to bypass and connect longer temporal dependencies, making it especially potent for traffic flow predictions with erratic and extended patterns. Another distinctive aspect is its non-standard 6-layer CNN, meticulously designed for in-depth spatiotemporal correlation extraction. The model comprises (1) the specialized CNN feature extraction, (2) the GRU-SKIP enhanced long-temporal module adept at capturing extended patterns, (3) a transformer module employing encoder-decoder and multi-attention mechanisms to hone prediction accuracy and trim model complexity, and (4) a bespoke prediction module. When tested against real-world datasets from California of Caltrans Performance Measurement System PeMS, specifically PeMS districts 4 and 8, the CNN-GRUSKIP consistently outperformed established models such as ARIMA, Graph Wave Net, HA, LSTM, STGCN, and APTN. With its potent predictive prowess and adaptive architecture, the CNN-GRUSKIP model stands to redefine ITS applications, especially where nuanced traffic dynamics are in play., Comment: 17 Pages, 18 Figures, 6 Tables

Assuming that a L\'evy-Driven Ornstein-Uhlenbeck (or CAR(1)) processes is observed at discrete times $0$, $h$, $2h$,$\cdots$ $[T/h]h$. We introduce a step-by-step methodological approach on how a person would verify the model assumptions. The methodology involves estimating the model parameters and approximating the driving process. We demonstrate how to use the increments of the approximated driving process, along with the estimated parameters, to test the assumptions that the CAR(1) process is L\'evy-driven. We then show how to test the hypothesis that the CAR(1) process belongs to a specified class of L\'evy processes. The performance of the tests is illustrated through multiple simulations. Finally, we demonstrate how to apply the methodology step-by-step to a variety of economic and financial data examples., Comment: 24 pages, 13 figures

The diversity of 2D materials and their van der Waals (vdW) stacking presents a fertile ground for engineering novel multifunctional materials and quantum states of matter. This permits unique opportunities to tailor the electronic properties of vdW heterostructures by the insertion of only a single 2D material layer. However, such vdW materials engineering at the atomic scale has yet to be investigated for spin-charge interconversion phenomena. Here, we report on the control of these effects at the monolayer level, where drastic increase in intensity and change in sign of THz spintronic emission are demonstrated by inserting a single layer of MoSe$_2$ between PtSe$_2$ and graphene in a fully epitaxial, large area stacked structure. By using a combination of spin and angle resolved photoemission and density functional theory to reveal the electronic and spin structures, we illustrate two different mechanisms relying on charge transfer and electronic hybridization for the formation of Rashba states, which are responsible for spin-charge conversion and hence the THz spintronic emission. These findings open new pathways to design, at the atomic scale, efficient THz spintronic emitters made of 2D materials and other spintronic devices based on spin-charge interconversion phenomena., Comment: 16 pages, 4 figures