The release of the Citizens Broadband Radio Service (CBRS) spectrum has enabled private network deployments, but signal strength in this band is highly sensitive to weather conditions. While traditional signal propagation models are often constrained by their assumptions and lack adaptability to varying environments, Machine Learning (ML) models offer a flexible alternative for predicting signal strength under complex conditions. In this work, we develop a Neural Network (NN) based predictive model for CBRS signal strength, leveraging a comprehensive set of weather parameters, collected over a 24month period from a live CBRS network in Buffalo, NY. Our model demonstrates a Mean Absolute Error (MAE) of 1.83 dB in predicting hourly signal strength, and we further evaluate its generalization capability by applying a transfer learning approach to another CBRS network in Elmira, NY, achieving a MAE of 2.13 dB. This study highlights the potential of ML methods in modeling weather-induced signal variability and their applicability across different network environments.

Healthcare data warehousing test automation is becoming a success through the help of AI-based technology that results in accuracy, efficiency, and data integrity on the automated test. The one central to personalized patient data that forms the core of traditional data warehousing solutions frequently faces problems of complexity and dissatisfaction. Deep learning with test automation solutions makes data ingestion, processing, and testing conversant through machine learning algorithms [1]. These systems encompass separate acts such as testing for data integration, testing the quality of data as well as performance testing removing the factors of time and human error. In healthcare organizations, this approach delivers the message of enhancing quality, compliance and organizational efficiency in value-based care and precision medicine initiatives. As such, artificial intelligence tools can be used to address the essential issues regarding data warehousing, and this can in turn enhance decision-making, and thus, enhance the health of patient and the overall population.

Evaluating privacy in data-driven software is hindered by its subjective nature and the multitude of privacyenhancing technologies. To address this, we propose a framework for categorizing data based on its privacy implications. This proposal aims to establish a standardized approach for quantifying privacy across various data sources, aligned with DevPrivOps.

The following paper aims to explore a number of projects that lie on the intersection of virtual reality, physical space and immersion, in particular that explore the concepts of dynamic immersions in teaching and learning spaces. Through this, the authors examine the application of Universal Design for Learning (UDL) templates in educational methodologies, particularly within higher education settings and specifically, in the context of how these affordances enhance inclusion and accessibility. The authors have conversely found limitations related to hardware form factors and interactive systems that cannot be customized to meet the needs of all users, resulting in marginalization. The authors wish to explore the manifold problems that lead to both intentional and unintentional exclusion ramifications, whilst also discussing relevant inclusion opportunities. Through this research, the authors aim is to present a topical overview of several use cases and to discuss how software applications can be better calibrated to meet the guidelines for UDL, and improve the accessibility of 3D and immersive interactions for innovative Virtual Learning Environments (VLEs). Note that in this context, VLE's can include bespoke spaces that are primarily intended to be virtual, immersive and also traditional pedagogical and didactic platforms. As part of this, the authors propose a new term: Extended Virtual Learning Environments (EVLEs) to encompass these developments, as well as immersive hardware innovations [1], [2].

AI chatbots have been reincarnated with the soul of generative AI, and now they are able to respond with intelligence and consciousness similar to that of humans. These AI chatbots that are based on generative AI are continuously evolving and enhancing their capabilities to act increasingly similar to humans. Generative AI represents all those AI technologies that can generate variety of contents including text, image, audio, video, code, virtual world, dataset, and web data. Different types of generative AI models have been developed to generate such contents, and Large Language Models (LLMs) and Large Multimodal Models (LMMs) are currently the most popular generative AI models used in most modern AI chatbots; where LLMs mostly handle and generate text data, whilst LMMs are able to handle and generate multiple data types or modalities such as text, image, audio, and video. The growing importance of AI chatbots based on generative AI necessitates an exploration of their capabilities in a detailed manner to obtain better understanding about their strengths and utilisations. Exploring the capabilities of AI chatbots that are based on generative AI, this paper will provide a systematic classification and thorough explanation of the capabilities of AI chatbots. It will classify the capabilities of AI chatbots into eight major categories: text generation and analysis, image generation and analysis, audio generation and analysis, video generation and analysis, code generation and analysis, virtual world generation and analysis, data (dataset) generation and analysis, and web data generation and analysis. This systematic classification and thorough exploration of the capabilities of AI chatbots will enable users to analyse these capabilities and the effective utilisation of AI chatbots in real life.

ChatGPT reshaped several aspects of AI technology and its usages in real life including searching, learning, reading, writing, and communicating using AI technology. It is powered with generative AI, and generative AI is an AI technology that is designed to process and generate different types of contents including text, image, voice, video, code, virtual world, dataset, and web data. Therefore, generative AI powered ChatGPT is able to process and generate different types of contents, and has the capability to assist humans in almost all areas of real life, and it is still expanding its capabilities regularly. This continuous expansion of the capabilities of ChatGPT requires a better understanding regarding its strengths and corresponding usages. Therefore, this paper will assess twenty capabilities of ChatGPT and provide an in-depth assessment, where it will highlight the strengths and limitations of each capability. This assessment of various capabilities will be based on the different types of contents it can generate and analyse: text, image, voice, code, virtual world, dataset, and web data. This indepth assessment of the capabilities of ChatGPT is focused on providing a better understanding of the strengths of ChatGPT and its corresponding usages based on these capabilities.

In high-speed vehicular communications, e.g., a high-speed train system, the beam between the base station and user is frequently misaligned, resulting in a huge amount of associated overheads. As a result, beam alignment is often challenging. This paper presents a new approach to beam alignment in millimeter wave high-speed train systems. The core idea of our approach is using Bayesian multi-armed bandit (MAB) in the beam search process, where we refer to the beams as arms. Furthermore, there exists an unimodal structure of the mean received signal strength (referred to as mean rewards) over the available beams (or arms). To this end, we develop an algorithm named Bayesian-Best Arm with Unimodality (B-BAU) that identifies the best arm with high confidence within a fixed-budget by exploiting the unimodal structure of the mean rewards over the arms. In the proposed method, by approximating the signal on each beam direction as a Gaussian random variable, for the Gaussian rewards with a Gaussian prior, we show that the expected probability that B-BAU misidentifies the optimal arm is upper bounded as O (1/ √), where is the budget. We then develop an algorithm for its frequentist counterpart named F-BAU. We demonstrate that F-BAU and B-BAU outperform the state-of-the-art algorithms through extensive simulations. While F-BAU is parameter-free, B-BAU performs better with the prior information.

Block-term tensor decomposition (BTD) was recently reconsidered in the context of array signal processing, and shown to be the natural choice in sensor arrays of increased dimensionality and channel multipath. Semi-blind joint channel estimation/data detection (JCD) was addressed in the special but important case of the uniform rectangular array (URA), via efficient algorithms that wed existing JCD schemes with BTD approximation. That work is revisited in this paper from a Bayesian standpoint to address the discrete nature of the input symbol constellation. This offers a new interpretation of the outstanding BTD JCD scheme among those previously developed. Moreover, two probabilistic methodologies are followed, which, when endowed with the BTD structure, result in additional JCD schemes that provide alternatives to our previous work in detection performance and computational complexity. The receivers are evaluated via simulation results at various noise levels, and different array and constellation sizes and favorably compared with the training-only-based solution.

Project management remains as one of the most popular post graduate management courses for international students at UK Universities. Students perceive or are advised that the skill set is highly transferable to a wide range of management opportunities and can be built upon almost any undergraduate discipline. However, many practitioners claim that the interconnectedness of the real life of projects can only ever be fully understood through on-the-job real life work experience. Hence the student dilemma, how to gain experience without experience? Project management simulation exercises attempt to bridge this gap. This research considers the effectiveness of a well proven simulation at a UK university comparing the applied knowledge of 17 different skill sets before and after the exercise through the lens of the student perspective. Furthermore, in depth semi-structured interviews with experienced tutors aim to decipher the mechanisms of learning so transferable principles can be prioritised for future exercise design. The study carried out during the 2020/21 academic year demonstrates that from a student perspective such educational approaches are highly effective with all the skill set analysed (a broad set of hard and soft skills) showing notable deltas before and after. In addition, students engage deeply with this method and find it fun and memorable which is important for future knowledge retrieval. The knowledge and experience of the tutors highlighted that game design was only 1 of 3 important elements crucial for learning to take place. Game management and the skill and motivation of the tutor have equal bearing on success. This research offers a conceptual framework to synthesis the findings and recommends that educationalists and employers collaborate and invest in future game development together rather than seek off the shelf solutions.

Frequency coupling between impedances causes misleading stability assessment through single coordinate Bode plots (dd− and qq− axes) for power electronics-dominated systems. To perform accurate stability analysis, it is crucial to incorporate this frequency coupling. Traditionally, the accurate stability of such coupled systems is analysed using a Multi-Input-Multi-Output (MIMO) framework, which involves employing the generalized Nyquist plot to assess the stability. However, if the coupling strength in these systems could be quantified and used to correct the diagonal elements of the MIMO system, the stability evaluation could be greatly simplified using single coordinate Bode plots. This paper introduces exactly that - a corrective factor based on a quantified frequency coupling parameter, known as coupling strength. This approach allows MIMO systems to be analysed within a Single-Input-Single-Output (SISO) framework. The corrective factor ensures simpler and precise analysis using single coordinate Bode plots. The effects of frequency coupling can be directly observed in the Bode plots of the diagonal elements of the corrected MIMO system matrix. Additionally, this method provides an accurate estimation of resonance frequencies and stability margins by direct inspection of the Bode plots. Simulation and experimental results are presented to validate the proposed stability assessment technique.

As Financial Technology (Fintech) continues to evolve rapidly, financial institutions recognise the need to integrate innovative solutions to enhance their risk management strategies. This acknowledgement stems from the potential benefits that Fintech offers in mitigating financial risks. However, despite this growing recognition, empirical research, particularly in emerging markets such as Jordan, remains scarce. To this end, this study aims to evaluate how Fintech applications impact the enhancement of financial risk management practices in Jordan. A questionnaire survey was developed and administered to 11 financial risk experts in Jordan to assess the effects of Fintech adoption on financial risk management. Participants evaluated the potential of Fintech applications, such as digital lending and credit, mobile banking, mobile payments, cryptocurrency, blockchain, and online trading, to improve financial risk management practices. The study findings indicate a perceived positive impact of Fintech adoption on enhancing the efficacy of financial risk management practices in Jordan. This study significantly contributes to the literature by shedding light on the positive impact of Fintech adoption on financial risk management, particularly in emerging markets like Jordan. Moreover, this study has important implications for policymakers, regulators, and banking practitioners, highlighting the importance of creating an enabling environment for Fintech adoption to address financial risks proactively.

A fundamental challenge in contemporary computational linguistics lies in the ability of language models to adaptively capture and maintain complex contextual relationships within diverse and dynamically shifting textual inputs. Transformer-Based Cascade Synthesis offers a pioneering solution to this problem through a cascading, multi-layered framework that progressively refines contextual embeddings at each stage, creating a more flexible, responsive model architecture capable of complex context handling across extended text sequences. Employing a structured series of transformations, the synthesis approach recalibrates embedding layers to achieve a balance between context sensitivity and computational efficiency, facilitating coherent, syntactically accurate language generation even in complex linguistic scenarios. Experimental results demonstrate substantial improvements in contextual coherence, lexical diversity, and syntactic precision, suggesting that Transformer-Based Cascade Synthesis achieves a marked enhancement in model adaptability and response fidelity compared to traditional static embedding methods. This methodology's effectiveness across diverse language tasks highlights its potential to set a new standard for adaptability in large language model architectures, contributing meaningfully to the broader pursuit of human-like language comprehension in artificial intelligence.

This paper presents channel measurements and modeling at 28 and 160 GHz in an industrial-style open office environment with angular information at the receiver. Channel impulse response snapshots are captured at 145 measurement positions for both frequencies and are evaluated in terms of power metrics, delay, and angular characteristics. Frequencydependent models for path loss, delay, and angular spread of arrival and K-factor are estimated, and large-scale parameter cross-correlations are evaluated. The results show that delay and angle spread are not frequency dependent in this environment although twice as many multi-path components were identified at 28 GHz compared to 160 GHz.

Imaging spectroscopy and lightweight unmanned aerial vehicles (UAVs) have revolutionized remote sensing of vegetation, by providing high spectral and spatial resolution data. Comprehensive wetland monitoring should be based on reliable mapping of biophysical aquatic vegetation parameters, which in turn requires low bias in measured spectra, e.g. minimization of reflectance anisotropy. This study aims to quantitatively investigate how sensor scan direction and solar zenith angle (SZA) affect vegetation spectra over two dominant aquatic plants, Phragmites australis and Nuphar lutea, representing different functional types (i.e. emergent and floating) with divergent canopy structure and affinity with water, using data collected with a hyperspectral (400-1000 nm range) push-broom sensor mounted on a UAV. Anisotropy factor (ANIF), mean reflectance scatterplots and regression analysis were used for assessing the magnitude of spectral anisotropy at both pseudo-leaf and canopy scales. Our findings showed more accentuated anisotropic behaviour in the visible than near-infrared domain at SZA<60° as the scan direction approaches that of solar principal plane. The increase in reflectance in near-forward direction affects the spectra of floating-leaved N. lutea at both leaf and canopy scales, suggesting the presence of water (as a film over leaves or as canopy background) as a likely anisotropy driver. Backward hotspot is evident in canopy spectra of emergent P. australis, with driving mechanisms similar to terrestrial species (i.e. shadow-hiding). At SZA>60°, reflectance anisotropy appears to be negligible, independent of functional type and scan direction. This research underscores the importance of considering sun-target-sensor geometry in remote sensing studies on aquatic vegetation.

A document by Mathis Schmieder. Click on the document to view its contents.

This research presents a network-based simulation approach to model the impacts of species extinction and habitat loss on ecological networks, with a particular focus on the Amazon ecosystem. By constructing weighted directed networks that represent species interactions, we identify keystone species through centrality measures and assess their critical roles in maintaining ecosystem stability. Using community detection algorithms, we uncover the modular structure of these interactions, revealing key patterns of interdependence within the network. Novel simulations of environmental disturbances—such as habitat fragmentation, patch loss, and species migration—further explore how these disruptions affect network structure, biodiversity, and overall ecosystem resilience. The findings offer valuable insights into ecosystem vulnerability and provide actionable guidance for conservation strategies aimed at mitigating biodiversity loss in fragile ecosystems.

Large-scale language models have demonstrated exceptional capabilities across diverse applications, yet often struggle to dynamically incorporate real-time knowledge, which is essential for generating responses with high contextual relevance and coherence. Addressing this limitation, the concept of Dynamic Knowledge Interpolation (DKI) is introduced as a transformative mechanism enabling seamless integration of external knowledge sources directly into the generation process, enhancing adaptive responses across various domains. By facilitating continuous and context-sensitive interpolation between static and dynamic information, DKI significantly improves performance in key areas, including perplexity, contextual coherence, and response accuracy, as evidenced by quantitative analyses and case studies. The DKI architecture, rigorously tested across multiple datasets, shows that adaptability in language models can be markedly enhanced without sacrificing robustness or scalability, thereby establishing DKI as an advanced approach to refining language model efficacy. Through substantial performance gains and an ability to maintain contextual awareness in dynamically evolving scenarios, DKI positions itself as an essential framework in the advancement of responsive and domain-sensitive language models.

Cardiovascular diseases (CVDs) account for approximately 17.9 million deaths annually, making them a leading cause of mortality globally. As the population increases, the early detection and treatment of heart diseases are becoming more challenging. With advancements in machine learning (ML), models can now predict heart diseases effectively. In this study, we aim to predict whether a person has cardiovascular disease based on 11 clinical features from the UCI heart failure prediction dataset. We compare the performance of five different machine learning algorithms: Gradient Boosting (GB), Random Forest (RF), K-Nearest Neighbors (KNN), Logistic Regression (LR), and Support Vector Machine (SVM). Our results show that SVM is the most efficient technique with an accuracy of 94.56%, which can aid doctors in reducing heart disease-related fatalities.