Reconstructing the mandible after the surgical removal of diseased sections is crucial for restoring functions such as chewing, speech, and facial integrity. However, achieving consistent bone union remains a significant challenge. This study hypothesizes that optimizing donor bone cuts, specifically through improved positioning and cut plane orientation, can enhance bone union outcomes. To test this hypothesis, we present a novel automated workflow that leverages Bayesian optimization. Applied to a mandibular body defect case, this approach integrates virtual reconstruction, remeshing, and biomechanical simulation, resulting in up to a 76% increase in predicted bone union propensity. Our findings indicate that donor cut planes warrant greater emphasis in surgical planning. Furthermore, automated optimization underscores the potential of this patient-specific, data-driven method to advance bone healing in mandibular reconstruction.

Access to suitable learning platforms is crucial for motivating learners and making Science, Technology, Engineering, and Mathematics (STEM) an attractive field of study. This study analyze the potential benefit of integrating Microcontrollers into STEM learning in the schools. The responses of participants in the STEM Outreach program is analyzed to evaluate the potential impact in encouraging the rural youths in taking up the STEM fields of study in higher education. Coding on two microcontrollers, namely the runlinc E32W controller and Arduino, was introduced in ten schools, involving 447 students, of which 211 were girls and 236 were boys. The focus was on making them understand and helping them project the larger scope of STEM through developing basic applications on automation, the Internet of Things (IoT), and robotics, particularly on solving local problems through technology interventions. Pre-and post-program empirical data were collected to examine how access to and hands-on learning could influence young learners' decision to pursue STEM as their chosen area of study. To gauge the impact of the hands-on learning experience, the researchers employed a Dependent Sample T-Test, ANOVA, Pearson correlation, and linear regression analyses to predict impact. All the tests indicated a significant positive impact on the influence of appropriate learning tools in STEM interest among the students. Further, the SPSS ANN algorithm predicted the significant impact on students' perception of STEM and its future scopes.

Recursive mechanisms have become essential in bridging the gap between linear processing capabilities and the intricate demands of extended textual coherence. Yet, traditional models often struggle to retain semantic and contextual clarity across evolving dialogues and multi-layered interactions, highlighting a pressing need for adaptive context handling frameworks. Introducing Recursive Context Layering (RCL), a novel approach that recursively structures context in Large Language Models (LLMs), this study provides a transformative methodology for retaining thematic alignment and managing complex discourse across extended text. Through recursive layering, RCL establishes a cumulative context system where each layer incrementally refines semantic depth, dynamically adapting to shifts in conversational focus while mitigating the challenges of semantic drift. The RCL framework allows LLMs to seamlessly adjust to varying levels of contextual granularity, maintaining response coherence in both structured and fluid language environments. Experimental findings substantiate the efficacy of RCL, demonstrating substantial improvements in semantic accuracy, contextual consistency, and retention metrics over baseline models, marking a significant leap in LLM design for tasks requiring sustained contextual depth and responsiveness. Overall, Recursive Context Layering not only enhances contextual fidelity in language processing applications but also sets a robust foundation for recursive strategies, positioning it as a crucial advancement for models handling complex and extensive discourse.

Change detection (CD) in remote sensing (RS) imagery plays a crucial role in Earth observation tasks using satellite data. However, due to the high similarity between the foreground and background in RS images and the insufficient utilization of difference information, current deep learning (DL)based CD methods still suffer from boundary ambiguity and false changes. To address these issues, we propose an Edge-Guided Difference Enhancement Network (EGDENet). First, multi-scale feature maps are extracted from a feature extractor. To explicitly delineate the boundaries of changed objects, we introduce the Edge Enhancement Module (EEM), which captures high-frequency edge information from bi-temporal feature maps at each scale. Then, a Dual-Branch Difference Enhancement Module (DBDEM) is designed to learn the difference between the foreground and background, guided by spatial and channel information, reducing the impact of irrelevant changes on the detection results. Additionally, a Multi-Scale Difference Feature Fusion Module (MSDFFM) is employed to enhance the robustness to multi-scale features, improving the internal completeness of the detected change regions. The experimental results demonstrate that our lightweight network effectively utilizes difference information and produces more accurate change region boundaries, outperforming state-of-the-art (SOTA) methods on four public datasets.

This paper presents a cost-effective and scalable hybrid methodology for evaluating retrieval-augmented generation (RAG) systems using specialized pretrained models and advanced metrics, designed for critical domains like healthcare and finance. LLM judge-based evaluation approaches are hindered by significant score inconsistencies across identical input runs (up to 35% variations) and high computational costs while traditional NLP approaches, overly reliant on entity or phrase matching, lack a multi-faceted perspective and fail to capture deeper semantic understanding. Our methodology addresses these challenges with a novel approach that combines an ensemble of fine-tuned pretrained models and advanced NLP techniques, ensuring consistent, reproducible evaluations across multiple dimensions, while being tailored for offline scenarios to eliminate reliance on internet-connected systems or proprietary LLMs, offering a cost-effective solution with high accuracy in assessing semantic relevance, factual correctness, and context adherence. To enhance reliability, the approach employs a robust weighted scoring system using harmonic means combined with PCA, adaptive, and entropy-weighting techniques for trustworthy and consistent evaluation enabling seamless integration with existing systems, continuous metric adaptation and domain-specific customization, ideal for high-stakes applications that demand rigorous quality assessments without relying on external APIs.

We propose a novel approach to active neuromorphic sensing to rapidly focus and artificially generate contrast to improve event-based space imaging. Space imaging is the collection of timely and accurate visual data of objects in the space environment and is essential for space-based operations and infrastructure. Recently, neuromorphic event-based vision sensors have been demonstrated as effective space imaging sensors with their high-speed contrast-driven operation. We demonstrate the performance of variable focus liquid lens driven artificial contrast generation as an active sensing technique. We demonstrate an improvement to the performance of an event-based space imaging system by cycling rapidly through an optical power range to incidentally pass through and detect an optimal focus range at each oscillation. We show that our approach can improve an event-based space imaging system's source detection, sensitivity, noise filtering, and localisation error. Furthermore, we demonstrate a solution to the critical problem of accurately and rapidly focusing the event-based camera for space imaging. By improving the performance of event-based space imaging systems, this sensor's unique capabilities can be further leveraged in critical space imaging applications such as international space traffic management efforts, and other optical domains as a new approach to difficult target detection problems.

Course evaluation is a critical component in higher education pedagogy. It not only serves to identify limitations in existing course designs and provide a basis for curricular innovation, but also to offer quantitative insights for university administrative decision-making. Traditional evaluation methods, primarily comprising student surveys, instructor self-assessments, and expert reviews, often encounter challenges, including inherent subjectivity, feedback delays, inefficiencies, and limitations in addressing innovative teaching approaches. Recent advancements in large language models (LLMs) within artificial intelligence (AI) present promising new avenues for enhancing course evaluation processes. This study explores the application of LLMs in automated course evaluation from multiple perspectives and conducts rigorous experiments across 100 courses at a major university in China. The findings indicate that: (1) LLMs can be an effective tool for course evaluation; (2) their effectiveness is contingent upon appropriate fine-tuning and prompt engineering; and (3) LLM-generated evaluation results demonstrate a notable level of rationality and interpretability.

Predicting the future trajectories of agents in dynamic, multi-agent environments remains a fundamental challenge, especially when models lack interpretability, an essential factor for safety-critical applications like autonomous driving. We propose the Scene-level Trajectory Prediction Transformer (STPT), a novel framework that integrates diffusion-based generative modeling with Kan network mechanisms to capture both spatial and temporal dynamics of agent-environment interactions. STPT leverages a recursive diffusion process that refines trajectory predictions over multiple time steps, explicitly accounting for uncertainty and inter-agent dependencies. Importantly, we introduce a Shapley value-based feature attribution technique tailored for diffusion models, quantifying the global and scenario-specific importance of features such as traffic signals and lane geometry at every stage of the prediction process. Extensive evaluations on benchmark datasets demonstrate that STPT not only surpasses state-of-the-art trajectory prediction methods in accuracy but also sets a new standard in real-time interpretability, making it particularly suited for deployment in safety-critical systems requiring both precision and accountability.

The cascading of power modules can generate multilevel voltage output to enhance power transmission capacity and quality. However, conventional cascaded converters need excessively high numbers of modules for high quality due to the only linear increase of levels with modules. Modules whose voltages following a decreasing sequence, e.g., a binary geometric progression (1, ½, ¼, …), imitate analog-digital conversion and generate very fine output shapes through a linear combination of modules with finer and finer steps. However, the use of modules with decreasing voltages rapidly reduces the energy content per module and typically requires dedicated module designs for every module voltage level. Thus, the large voltage differences thwart the initial goal of a uniform, modular, and readily scalable circuit. Furthermore, keeping this balance with very different voltages and energy contents in the same string becomes challenging, while modules with smaller voltages need to switch by orders of magnitude faster than larger ones. Active or diode-enabled parallel modes that simplified balancing and increased utilization in conventional cascaded circuits appear not helpful across the large voltage differences. This paper proposes a novel asymmetric converter design featuring a very different sequence of module voltages, an optimal low-cost control scheme, and accurate sensorless module-voltage regulation. The circuit does not operate any module below half the highest-voltage module (dc link voltage). Compared to previously suggested asymmetric cascaded circuits, modules will have significantly closer rated voltages, yet their voltages never decay to negligible values at higher voltage levels. The proposed tighter rated voltage range allows for a uniform module design with fully interchangeable modules. The reduced voltage difference between neighboring modules exponentially reduces the total size and dimensions of the inductors required to limit the balancing current. Moreover, we introduce an optional pseudo-parallel mode for efficient module-to-module energy exchange. Simulations using MATLAB/Simulink and a hardware prototype confirm the effectiveness of this approach.

Artificial General Intelligence (AGI) requires an infrastructure, for algebraically expressing composable structures and operations, built on powerful abstractions. In a well-trained system built on these principles, learning to make appropriate algebraic substitutions would be sufficient to make intelligent responses. The power of Category Theory (CT) to dynamically compose appropriate granular units of computations and Homotopy Type Theory (HoTT) to help with semantics and validate the compositions, is necessary for this. This paper describes the roles CT and HoTT could play in providing a mathematical foundation for implementing a framework suggested by the Subjective Emergence of Objective Minds (SEOM) model developed earlier by the author. The paper concludes by observing that insights derived here, regarding Deep Neural Networks (DNN) and other related approaches to artificial intelligence, are necessary to put us on the correct path to AGI.

The unprecedented demand for seamless multimedia services necessitates novel approaches to optimize caching and computing. This paper presents an innovative framework using digital twin (DT) technology to create a dynamic, virtual representation of multimedia service infrastructure, enabling real-time optimization for user-centric applications. By leveraging DT's predictive capabilities and integrating them with advanced caching and computing strategies, the proposed system improves resource allocation, reduces latency, and enhances Quality of Experience (QoE) for users. Our simulation results demonstrate significant performance improvements over traditional approaches, establishing the potential of DT-based systems in next-generation multimedia services.

A pure weight-based test case generator for context-free grammars is introduced, where weights are set for each rule and used as unique driver to identify which one to apply during the derivation process; we demonstrate why static flat weights generally cannot work whereas, introducing a simple penalty mechanism, a dynamic weight-based approach is always able to produce random but finite sentences; then, looking at the source grammar from an original perspective, we proceed towards the definition of constraints through which, while still guaranteeing the convergence of the derivation, static but balanced weights can be identified: balanced weights own the interesting capability to drive the generation of an inexhaustible number of extremely random, structurally complex but finite sentences by an automatic test case generator having a dramatically simple derivation logic. Thereafter, we illustrate how to design a weight-based system which, by exploiting the characteristics of balanced weights, can be easily configured to produce sentences that address any specific coverage metric, including those not having a definite generative strategy yet. A complete example of the application of the balanced concept and the weight-based system is then included and experimental results are eventually shown.

In this paper, we discuss the manufacture and testing of a modular snake inspired robot, designed with the specific objective of minimising actuation requirements to enable serpentine movement, whilst still permitting dorso-ventral motion. To achieve this, we use ball gears to increase the range of motion at joints that are connected to the robotic snake body by linkages. While we use different motors to enable both dorso-ventral and lateral kinematics, the research presented here focusses on the lateral (serpentine) kinematics. The serpentine kinetics of the robotic snake are mapped for both linear and non-linear trajectories. We find that despite our robot's use of simple actuation mechanisms, serpentine movements leading to forward motion are possible. The paper instigates new scope for the use of ball gear mechanics to simplify serpentine movement in modular robots, whilst still permitting dorso-ventral kinematics through the same joints.

The increasing complexity of user interactions requires language models that can dynamically adapt to complex contextual cues. Adaptive Contextual Synthesis (ACS) introduces a multi-layered framework that enhances semantic alignment by recalibrating responses in real-time according to the intricacies of input prompts. Through integrating adaptive feedback loops and probabilistic adjustments, ACS refines the coherence and relevance of generated text, addressing limitations in existing models' contextual adaptability. Empirical evaluations demonstrate that ACS significantly improves semantic precision and user satisfaction, showing its potential to advance the development of more contextually aware language models.

The fifth-generation new radio coding techniques, such as low-density parity-check (LDPC) codes, have made significant progress. However, their limitations in the low error floor region prevent them from meeting the stringent bit error rate (BER) targets required for sixth-generation (6G) communications. 6G aims to support services with BER as low as 10 −9 , providing the reliability needed for mission-critical applications. his is particularly crucial for ultra-reliable lowlatency communication. In this paper, we present a two-stage decoding method that combines the strengths of layered LDPC decoding and federated transfer learning (FTL). The layered LDPC decoding stage effectively handles the challenging waterfall region, reducing errors during the initial decoding process. In addition, the FTL model adapts to mitigate the error floor and accelerate its decline. This combination achieves lower error rates and faster performance improvements, reducing the need for retransmissions and enhancing overall system efficiency.

Hidden Markov Models (HMMs) are fundamental tools for modeling sequential data, yet traditional methods for evaluation and estimation often struggle with computational inefficiency, especially when applied to large datasets. In such cases, the evaluation of sequence likelihoods becomes imprecise, impairing the performance of the Baum-Welch algorithm, which relies on these likelihood calculations to propose transition and emission matrices. In this work, we present two innovative techniques aimed at improving computational efficiency in these processes. For model evaluation, we introduce a non-parametric, sample-based method that can be used to compare likelihoods using small, randomly sampled segments of data, thereby reducing computational load while preserving statistical integrity. For parameter estimation, we leverage a novel mathematical relationship between transition and emission matrices and observed transitions, which constrains the search space and eliminates the need for iterative likelihood computations. Results from Monte Carlo simulations demonstrate that our methods substantially enhance computational performance without compromising accuracy.

Preprocessing methods are important in enhancing prediction performance for time-series administrative data. This study underscores the importance of preprocessing methods by comparing two data representation techniques, that can be used with sliding window techniques for time-series administrative healthcare data, for use with gated recurrent unit (GRU) networks. The first method uses a sequence event representation and the second employs a sequence matrix representation. The evaluation was conducted through a retrospective administrative healthcare data case study to predict multiple outcomes. The target outcomes were the first instances of healthcare encounters indicating homelessness or police interaction that appeared in the healthcare data. Results reveal that the GRU combined with sequence matrix representation and sliding window outperformed the sequence of events with the sliding window by over 20% in the area under the curve (AUC) and sensitivity for both outcomes. Given the data used in this analysis, sequence matrix representation was superior to sequence event representation while using GRU models. The results remained consistent across the evaluation in real-world clinical frameworks using two trigger methods for prediction time, such as the sliding window and clinical demand window structures.

Understanding speech in noisy environments remains a critical challenge for hearing-loss intervention development. Advancement in methods for enhancing speech recognition necessitate extensive datasets for training and testing algorithms. The majority of speech corpora focus on individual utterances, dyadic interactions and scripted recordings (e.g. of phonetically balanced sentences), despite the fact that much of our social communication occurs in dynamic, small group interactions. This paper presents a systematic review of existing speech corpora that encompass both audio and visual recordings of small group conversations. In doing so, we examine the currently available data for testing and evaluating computational audio-visual speech enhancement models. The available datasets for testing speech enhancement techniques are varied and often specific to set use cases. This results in restrictions in technology development due to limitations imposed by the available training and testing databases. There is a need to develop a corpus of audio-visual multi-talker speech that can be used to train audio-visual speech enhancement technology and further cognitive hearing-related research. Existing audio-visual corpora are typically confined to the specific paradigms for which they were collected. Studies utilising speech with various types and levels of background noise generally achieve this by combining datasets containing utterances or dialogue with databases of background noise. Existing 2 available datasets do not incorporate free-flowing speech between multiple talkers whilst also accounting for differing listening environments and hearing abilities. No database exists to computationally explore the needs and preferences of hearing aid users in typical situations such as conversations in social environments or while navigating busy events. We propose the generation of a new dataset containing audio and visual information from multiple talkers interacting spontaneously in a variety of sound environments. Such a corpus would provide audiovisual data of free-flowing conversations to evaluate and fine-tune technological developments in realistic scenarios. Such evaluation data is necessary to develop speech enhancement models that will be able to cope with the complex demands of hearing in the real world.