RSS 1.0
The system will be going down for regular maintenance. Please save your work and logout.
Computer and Engineering Sciences
Recent Submissions
-
Long time stability and strong convergence of an efficient tamed scheme for stochastic Allen-Cahn equation driven by additive white noiseHuang and Shen [Math. Comput. 92 (2023) 2685–2713] proposed a semi-implicit tamed scheme for the numerical approximation of stochastic Allen–Cahn equations driven by multiplicative trace-class noise. They showed that the scheme is unconditionally stable on finite time intervals and can be efficiently implemented. In this paper, we investigate the long-time stability of this tamed scheme for stochastic Allen–Cahn equations driven by additive white noise. We also address the strong convergence analysis of the associated fully discrete scheme within the Galerkin finite element framework. The main contributions of this work are as follows: (i) by constructing a suitable Lyapunov functional, we establish the unconditional long-time stability of the tamed method; (ii) we rigorously derive the strong convergence rates of the fully discrete scheme obtained by coupling the tamed approach with the finite element method. Numerical experiments are provided to validate the theoretical analysis and demonstrate the effectiveness of the proposed scheme.
-
Evaluating dialogue adaptability: A comparative study of self-feeding mechanisms in federated and centralized chatbot architecturesEvaluating chatbot adaptability after deployment remains critical for ensuring ongoing relevance and user satisfaction. While previous research compared federated vs. traditional architectures for intent classification, the post-deployment adaptation capabilities of chatbots-particularly through self-feeding mechanisms-remain relatively unexplored. This paper evaluates self-feeding mechanisms in federated and centralized chatbot architectures, specifically investigating the impact of explicit and implicit user feedback on chatbot adaptability post-deployment. We empirically assess the effectiveness of these feedback loops in addressing data drift and improving intent classification accuracy over time. Through a comparative analysis, the study highlights distinct strengths and limitations in each approach, providing new insights into how chatbots can continuously enhance user experience and learning performance. Our findings emphasize the critical role of self-feeding mechanisms for sustainable chatbot operations, extending beyond initial training toward robust, ongoing performance improvements, complementing the literature on privacy-centric federated chatbot systems.
-
Packing, BiLSTM, and attention for privacy-preserving intent classification: Practical upgrades across centralized and IID-federated learningWe introduce a deployment-oriented intent classification framework that delivers a strong accuracy–efficiency–calibration trade-off without external pretraining or any changes to the data: an attentional BiLSTM coupled with four training/architecture elements: length-aware sequence packing, bidirectional recurrence, variational (locked) dropout across time, and attention pooling. The same compact architecture is evaluated in two regimes: centralized learning (CL) and federated learning with IID client partitions (IIDFL) using FedAvg, with shared hyperparameters to isolate the impact of the modeling recipe. The pipeline produces publication ready artifacts (CSV logs, learning curves, per-class F1, reliability diagrams with ECE, and round-wise confusion matrices for FL) to enable transparent, reproducible assessment. On a multi-intent dataset representative of production constraints, the model attains high Accuracy and Macro-F1, improved tail robustness (higher worst-class F1), and low Expected Calibration Error in CL; under IID-FL it exhibits smooth round-wise convergence toward the centralized reference while maintaining a modest communication budget per round (one broadcast plus m client uploads with 32-bit floats). This work contributes: (1) a principled, portable LSTM recipe: (a) packing, (b) BiLSTM, (c) locked dropout, and (d) attention—that improves recognition and calibration without additional data; 2) an IID-FL evaluation with round-wise diagnostics and communication estimates; and (3) a reference implementation that outputs all metrics and figures needed for rigorous, deployment-focused reporting in privacy-conscious assistants.
-
FireNet-Lite: A separable convolutional network for ultra-efficient fire image classificationFire outbreaks pose serious threats to ecological systems, human lives, and property, necessitating rapid and efficient detection mechanisms. While traditional fire detection approaches using deep learning have shown promising results, many rely on computationally intensive architectures, limiting their applicability in real-time scenarios and resource-constrained environments. In this study, we propose FireNet-Lite, a lightweight convolutional neural network optimized with depth-wise separable convolutions for fire image classification. FireNet-Lite achieves a test accuracy of 93% and a high recall of 96.57%, using only 7,693 parameters, thereby significantly reducing computational overhead. Owing to its compact size and low latency, the model is well-suited for deployment on edge devices such as surveillance cameras and drones, where efficient and timely fire detection is critical. Experimental results demonstrate that FireNet-Lite effectively balances detection performance with computational efficiency, outperforming many existing models in this regard. Furthermore, the model exhibits robustness to variations in lighting conditions, backgrounds, and flame intensity, enhancing its reliability in diverse real-world environments. Its architecture is specifically designed for low-latency inference, making it highly applicable in real-time fire detection systems. The promising results of FireNet-Lite highlight its potential as a scalable, practical solution for next-generation fire prevention technologies.
-
Developing sensor technology and algorithm for enhancing accuracy of monitoring anomaly sleep at homeSleep apnoea represents one of the most prevalent respiratory disorders worldwide, affecting approximately 49.7% of men and 23.4% of women. While polysomnography (PSG) remains the gold standard for diagnosis, its complexity, high cost, and limited accessibility have motivated the development of alternative diagnostic approaches. As respiration and blood oxygen saturation (SpO2) serve as key indicators in sleep apnoea assessment, photoplethysmography (PPG) has emerged as a promising technology for non-invasive monitoring. This dissertation presents a comprehensive investigation of PPG-based sleep monitoring systems, encompassing hardware development, signal processing algorithms, and advanced machine learning techniques for abnormal sleep detection. The research establishes a systematic framework for wearable sleep monitoring through three key phases of development and validation. First, a non-invasive continuous monitoring system was developed to assess sleep apnoea via SpO2 and heart rate (HR) measurements. Various breathing experiments were performed to investigate the relationship between breathing patterns, SpO2 fluctuations, and HR variations during apnoea events. Motion artifact (MA) analysis revealed directional and frequency dependencies, various adaptive filters were implemented to compare their effectiveness in MA removal. Next, an integrated framework combining variational mode decomposition (VMD) denoising with deep learning autoencoders was developed for automated abnormal SpO2 detection. The study investigated both the relationship between architectural complexity and performance gains, as well as the trade-offs between model complexity and computational requirements. The performance of deep learning autoencoders was evaluated across datasets with different characteristics. The VMD algorithm achieved an SpO2 root-mean-square error (RMSE) of 0.862% against clinical references. Among four autoencoder architectures, the long short-term memory (LSTM) autoencoder performed best (accuracy: 0.86) by capturing temporal SpO2 dynamics during apnoea events, though with higher computational costs. Finally, the feasibility of direct raw PPG signal analysis was investigated, bypassing error-prone SpO2 calculations. A wireless multi-sensor system integrating PPG and accelerometer data with Bluetooth connectivity was developed and evaluated. Comparative analysis of nine machine learning and deep learning models revealed that neural network architectures consistently outperformed traditional methods, with the novel 1D-Branch CNN achieving optimal balance between accuracy (0.92) and computational efficiency (~25% faster than CNN-LSTM). This system demonstrates the feasibility of using raw PPG signals for reliable abnormal sleep detection in clinical applications, emerges as particularly promising, offering an effective compromise between the computational efficiency, high performance and the need for generalization in variable environments.
-
Dynamic pricing-driven load optimization in islanded microgrid for home energy management systemsIn order to maximize residential energy use and reduce electricity costs, home energy management systems (HEMS), are crucial. A distinct dynamic pricing-driven load optimization technique for HEMS running in islanded mode where grid access is either limited or nonexistent is presented in this paper. Incorporating important distributed energy resources like photovoltaic (PV) systems, electric vehicles (EVs), battery energy storage systems (BESS), and limited grid access, the optimal schedule is also used for real-time dynamic pricing analysis. A planned short-duration outage and a full 24-hour islanding scenario are the two different outage scenarios that are assessed. In short-duration outages, grid dependency decreases to roughly 10% in the spring and approximately 50% in the summer and winter, according to a thorough seasonal analysis conducted in the spring, summer, and winter. The suggested approach guarantees total self-sufficiency in the 24-hour outage scenario, with local resources satisfying load demands in full.
-
Banal deception and human-AI ecosystems: A study of people’s perceptions of LLM-generated deceptive behaviourLarge language models (LLMs) can provide users with false, inaccurate, or misleading information, and we consider the output of this type of information as what Natale calls ‘banal’ deceptive behaviour [53]. Here, we investigate peoples’ perceptions of ChatGPT-generated deceptive behaviour and how this affects people’s behaviour and trust. To do this, we use a mixed-methods approach comprising of (i) an online survey with 220 participants and (ii) semi-structured interviews with 12 participants. Our results show that (i) the most common types of deceptive information encountered were over-simplifications and outdated information; (ii) humans’ perceptions of trust and chat-worthiness of ChatGPT are impacted by ‘banal’ deceptive behaviour; (iii) the perceived responsibility for deception is influenced by education level and the perceived frequency of deceptive information; and (iv) users become more cautious after encountering deceptive information, but they come to trust the technology more when they identify advantages of using it. Our findings contribute to understanding human-AI interaction dynamics in the context of Deceptive AI Ecosystems and highlight the importance of user-centric approaches to mitigating the potential harms of deceptive AI technologies.
-
Correction to: Visualization for epidemiological modelling: challenges, solutions, reflections and recommendations (2022) by Dykes et al.In the original version of this article, references 113–120, 123–140 and 143 were incorrectly numbered. This has been corrected on the publisher’s website.
-
Same structures, different settings: exploring computing capital and participation across cultural contextsThe number of people choosing to study computing in higher education remains low. Previous research has developed a research instrument to identify factors underlying student participation grounded in Bourdieu’s sociocultural theory. This study replicates and extends the original study, which identified key social, cultural, and psychological factors linked to computing education participation in Sweden. Using the validated research instrument, we distributed a survey across 11 UK universities, gathering responses from 131 students. Through Confirmatory Factor Analysis, we assessed the robustness of the original study’s constructs — career interest, subject-specific interest, influence from family and friends, confidence, and sense of belonging — and their relationship to subject choice in computing. After model refinements, the replication confirmed and validated the factor structure, supporting the stability of these constructs and their relationship to computing subject choice across cultural contexts. In addition, the current study adds additional open-ended questions to the research instrument to help explain the quantitative results. A thematic analysis further explains the correlation between previous experience, social influence, confidence, and gender, and how that relates to participation in the field. By replicating and extending the original study’s methodology, this research evaluates the reliability and generalisability of its conclusions, contributing to the evidence base needed to design interventions that broaden participation in computing education.
-
The effect of multiplayer game modes on inter-player data for player experience modellingResearch into social compliance, emotional contagion and behavioural synchronicity shows promise for various avenues of work concerning human-computer interaction, and a wider understanding of emotion. Despite their relevance, few studies have applied findings from these domains to player experience modelling in a multiplayer game, in itself having applications in entertainment, education and healthcare. Further to this, of the little work making use of inter-player data to model aspects of player experience, none considers the differences that may be found across common multiplayer game modes. This work therefore makes use of data collected across players in a series of common multiplayer game modes, considering the utility of inter-player data for predictive modelling using artificial neural networks in each. Results suggest that approaches modelling measures of players' experiences in terms of discrete emotion intensities are best made using their own facial expressions in nearly all circumstances, but past this, facial expression data from team based and competitive game modes shows the greatest promise. Considering the additional data separations available to team-based gameplay, we find that data collected from players on an opposing team shows greater utility for prediction of target player experience than data collected from a player on the same team. Regarding this, we make suggestions for the most applicable avenues for future research into the utilisation of inter-player data for emotional modelling.
-
Study on modelling and optimising controllers for heating systems in buildingsThe UK, like the rest of the world, is working toward net-zero carbon emissions by 2050. A significant contributor to national emissions (17%) is domestic space heating, making efficient building design and retrofitting crucial. This thesis presents methodologies for simulating, controlling, and analysing domestic heating systems to reduce energy use while maintaining thermal comfort. The core framework, Inverse Dynamics-based Energy Assessment and Simulation (IDEAS), was enhanced into IDEAS+ to better align with the UK’s Standard Assessment Procedure (SAP) for building regulations. Key improvements include: • A new thermal comfort algorithm for more accurate modelling of human heat perception and support for niche heating systems. • Dynamic free heat gain calculations, improving precision and SAP compliance. • An updated optimum start method to optimize heating schedules based on system capacity. IDEAS+ was first calibrated using direct electric heating, then applied to Gas Condensing Boilers (GCBs) and Air Source Heat Pumps (ASHPs). To further reduce emissions, optimizing control systems for dynamic energy markets was explored. Traditional optimization methods are slow, requiring iterative simulations. Instead, this thesis introduces a new method - OPTimal Inverse Control (OPTIC), which embeds cost-function optimization directly into IDEAS+’s inverse dynamics control. This allows real-time optimization alongside system operation, improving performance. OPTIC was tested with battery storage, dynamically adjusting to fluctuating energy prices and carbon intensity while ensuring thermal comfort. Two case studies demonstrated simulation based and practical based applications. One was a block of flats in Eastbourne, modelled in IDEAS+ for retrofit analysis, then simulated with a heat pump network and OPTIC-controlled storage as a simulation only based study. Whilst, the other was a commercial property with PV arrays, heat pumps, and battery storage, controlled via an Industrial PC running OPTIC-based C++/C# code at the property in a practical application. These methods provide scalable solutions for reducing emissions in residential and commercial heating systems, supporting the UK’s net-zero targets.
-
Designing Value-Aligned Traffic Agents through Conflict SensitivityAutonomous traffic agents (ATAs) are expected to act in ways tat are not only safe, but also aligned with stakeholder values across legal, social, and moral dimensions. In this paper, we adopt an established formal model of conflict from epistemic game theory to support the development of such agents. We focus on value conflicts-situations in which agents face competing goals rooted in value-laden situations and show how conflict analysis can inform key phases of the design process. This includes value elicitation, capability specification, explanation, and adaptive system refinement. We elaborate and apply the concept of Value-Aligned Operational Design Domains (VODDs) to structure autonomy in accordance with contextual value priorities. Our approach shifts the emphasis from solving moral dilemmas at runtime to anticipating and structuring value-sensitive behaviour during development.
-
Designing value-aligned traffic agents through conflict sensitivityAbstract Autonomous traffic agents (ATAs)-automated systems with high level of autonomy in traffic environments must not only guarantee safety but also act in accordance with legal, social, and moral values. In this short version, we adopt the epistemic game-theoretic conflict model of Damm et al to characterise value conflicts-situations where competing, value-laden goals cannot all be satisfied. As a mean to align the decision making of an ATA with stakeholder preferences, we introduce Value-Aligned Operational Design Domains (VODDs). They represent autonomous decision making scopes that guide an agent's conflict resolution and specify handover rules
-
From data-compliance to model-introspection: Challenges in AV rule compliance monitoringAutonomous vehicles (AVs) are expected to comply with traffic laws, ensure safety, and provide transparent explanations of their decisions. Achieving these goals requires monitoring architectures that pro- cess large volumes of sensor, control, and contextual data. While real-time perception and decision-making are functionally indispensable, storing and using this data for auditing or improvement raises unresolved legal and technical challenges. Data protection regulations—such as the GDPR—mandate that personal data processing be limited to what is strictly necessary for specified purposes (Art. 5(1)(b), (c), and (e)). Yet, in practice, what counts as “necessary” remains ambiguous. This tension gives rise to the data-justification gap: the lack of systematic methods to determine which logged data is both sufficient to support compliance assessments and minimal under data protection constraints. At the same time, aligning formalized rules with their legal intent poses a separate but interrelated challenge—the alignment problem. Legal norms are often ambiguous or context-dependent, and existing monitoring frameworks rarely guarantee that formal specifications faithfully reflect legal meaning. This paper outlines a research agenda for bridging these gaps. We propose an integrated approach com- bining formal methods, legal reasoning, and runtime monitoring to develop data-justification frameworks. Such frameworks would enable developers to generate interpretable rule formalizations, synthesize minimally sufficient monitors, and justify data collection in a transparent and legally defensible manner.
-
Efficient Spectrum Sharing in Cognitive Radio Networks With NOMA Using Computational IntelligenceThe integration of Cognitive Radio Networks (CRNs) with Non-Orthogonal Multiple Access (NOMA) offers great potential for improving spectral efficiency in 5G and Beyond-5G (B5G) networks. This study proposes an efficient spectrum-sharing technique for dual-hop CRNs using NOMA, optimized by an Improved Artificial Bee Colony (IABC) algorithm and guided by a Single Input Single Output Fuzzy Rule-Based (SISO-FRBS) System. In this setup, a distant primary transmitter communicates with the primary receiver via a secondary NOMA relay. The objective is to maximize the sum data rate of secondary users (SUs) while minimizing total transmission power. SISO-FRBS enhances IABC search process by dynamically guiding the search agents, improving both optimization quality and convergence. Simulation results show that the proposed scheme achieves the primary data rate benchmark of 5bit/s/Hz at a transmit power of 19mW, compared to 23mW with traditional ABC, achieving a 19.04% improvement in power efficiency.
-
MSAF: A cardiac 3D image segmentation network based on Multiscale Collaborative Attention and Multiscale Feature FusionAccurate segmentation of cardiac structures is essential for clinical diagnosis and treatment of cardiovascular diseases. Existing Transformer‐based cardiac segmentation methods mostly rely on single‐scale token‐wise attention mechanisms that emphasize global feature modeling, but they lack sufficient sensitivity to local spatial structures, such as myocardial boundaries in cardiac 3D images, resulting in ineffective multiscale feature capturing and a loss of local spatial details, thereby negatively impacting the accuracy of cardiac anatomical segmentation. To address the above issues, this paper proposes a cardiac 3D image segmentation network named MSAF, which integrates Multiscale Collaborative Attention (MSCA) and Multiscale Feature Fusion (MSFF) modules to enhance the multiscale feature perception capability at both microscopic and macroscopic levels, thereby improving segmentation accuracy for complex cardiac structures. Within the MSCA module, a Collaborative Attention (CoA) module combined with hierarchical residual‐like connections is designed, enabling the model to effectively capture interactive information across spatial and channel dimensions at various receptive fields and facilitating finer‐grained feature extraction. In the MSFF module, a gradient‐based feature importance weighting mechanism dynamically adjusts feature contributions from different hierarchical levels, effectively fusing high‐level abstract semantic information with low‐level spatial details, thereby enhancing cross‐scale feature representation and optimizing both global completeness and local boundary precision in segmentation results. Experimental validation of MSAF was conducted on four publicly available medical image segmentation datasets, including ACDC, FlARE21, and MM‐WHS (MRI and CT modalities), yielding average Dice values of 93.27%, 88.16%, 92.23%, and 91.22%, respectively. These experimental results demonstrate the effectiveness of MSAF in segmenting detailed cardiac structures.
-
Inter-player data for the prediction of emotional intensity in a multiplayer gameThis work assesses the feasibility of predicting emotional intensities for a given player in a testbed multiplayer game, using facial expression data collected from other players in the multiplayer group. Whilst there is significant literature on the utilisation of affect detection to build models of player experience, little research considers the additional data provided from other players in a multiplayer setting, despite the inherently shared experiences that they provide. A dataset describing 24 participants is collected, detailing ten levels of a testbed game, Colour Rush, with data collected describing facial expression activity and responses to the Discrete Emotions Questionnaire. The viability of modelling uncaptured player experiences is tested using artificial neural networks trained on facial expression data from target players, non-target players and a combination of both. Findings indicate that multiplayer data can be beneficial in the prediction of a target player’s emotional responses, although this holds true only in a minority of cases, and for specific groups of players.
-
Sensorimotor Synchronisation and Entrainment in Musical Timekeeping: Metronome Configurations and Preliminary Implications for Music EducationTimekeeping whilst playing music is a skill all musicians, especially drummers, require. Following a review of the literature on this subject, this thesis explores methods of measuring timekeeping accuracy of individuals and groups, and offers recommendations for approaches for future training. Using equipment and techniques accessible to musicians and non-musicians alike, the researcher has investigated how individual timekeeping, specifically the measure of Inter-tap Interval (ITI), is influenced by the presence, absence, and reintroduction of metronomes of various designs. The thesis also investigates how the influence of these differing metronome states interact with tempo and with the type of metronome (audio and visual). Similarly, the dynamics of group timekeeping and the interaction (or entrainment) between individuals in the group is also investigated. Participants were asked to report their perception of their individual performances under the different conditions of the experiments. The results show that tempo influenced the accuracy of timekeeping and the presence, absence and reintroduction of the metronome also had effects on accuracy. Individuals thought their timekeeping to be more accurate when the metronome was present and that they performed better as individuals as opposed to being part of a group. Detailed analysis of the results showed that the reintroduction of the metronome proved to have a significant effect on average ITI produced by participants, as did tempo. Metronome type had no significant influence on ITI in an individual or group setting. In the conclusion of the thesis, the author provides recommendations for future assessment and training of musicians in the skill of timekeeping, with respect to the measure of ITI.
-
Sustainable manufacturing of a Conformal Load-bearing Antenna Structure (CLAS) using advanced printing technologies and fibre-reinforced composites for aerospace applicationsConformal load-bearing antenna structures (CLAS) offer significant advantages in aerospace by reducing drag and weight through highly integrated designs. However, challenges remain in manufacturing, as traditional PCB methods create discontinuous arrays, while directly printed antennas on flexible substrates often lack mechanical strength. Additionally, neither approach integrates well with fibre-reinforced composites, which are widely used in modern aircraft. To address this, the next generation of CLAS must employ continuous surface substrates to maintain aerodynamic profiles and embed antenna systems within composite structures. This research introduces an innovative CLAS manufacturing method that integrates inkjet-printed silver nanoparticle antennas with composite fabrication. The antenna is printed onto Kapton film, which is then co-cured with woven glass fibre composites to ensure mechanical robustness and compatibility with aerospace materials. Flat and 100mm curvature samples were fabricated to investigate electromagnetic performance, with curvature effects analysed. Results confirm that the proposed method achieves both reliability and sustainability, producing smoothly curved CLAS with embedded antenna elements. However, frequency shifts and impedance mismatches were observed, attributed to discrepancies in dielectric constants and substrate volume variations. The conformality study revealed that curvature lowers resonant frequencies due to extended effective electric fields. This research establishes a promising CLAS fabrication approach, integrating sustainable printing with composites. The findings provide a benchmark for future conformal antenna studies and support industry-level advancements in high-integration aerospace antenna systems.
-
FireLite: Leveraging Transfer Learning for Efficient Fire Detection in Resource-Constrained EnvironmentsFire hazards are extremely dangerous, particularly in sectors such the transportation industry where political unrest increases the likelihood of their occurring. By employing IP cam eras to facilitate the setup of fire detection systems on transport vehicles losses from fire events may be prevented proactively. However, the development of lightweight fire detection models is required due to the computational constraints of the em bedded systems within these cameras. We introduce ”FireLite,” a low-parameter convolutional neural network (CNN) designed for quick fire detection in contexts with limited resources, in answer to this difficulty. With an accuracy of 98.77%, our model—which has just 34,978 trainable parameters—achieves remarkable performance numbers. It also shows a validation loss of 8.74 and peaks at 98.77 for precision, recall, and F1-score measures. Because of its precision and efficiency, FireLite is a promising.
