Mathematics and Computer Science: Research Updates Vol. 8

Comparative Models of Newcomer Integration for Organisational Adoption

2025-11-10T08:45:45+00:00

Strategies that address the needs of both the newcomer and the organisation can ease this transition and facilitate a smooth assimilation process. Welcoming and integrating newcomers is a critical challenge for organisations seeking to maintain a cohesive and productive workforce. This study utilises a comparative study to explore and contrast various models for newcomer integration within organisational contexts. This study aims to provide valuable insights into the comparative effectiveness of different integration models and offer practical recommendations for organisations seeking to improve their onboarding processes. The authors employ a comparative study methodology by conducting a systematic literature review. This approach allows for in-depth insights into the effectiveness of each onboarding model. The research identifies key components, outcomes, and challenges associated with each model. Through a detailed comparative analysis of models such as the Developers Joining Model, Onion Model, Identity Socialisation Model, Four C's Model, the characterisation framework, the onboarding types, and Traditional Onboarding Process, the study evaluates their performance across critical factors, including onboarding plans, mentorship, feedback, cultural fit, flexibility, and role clarity.

The findings reveal distinct strengths and limitations for each model, highlighting their varied effectiveness in promoting employee engagement and retention. The Identity Socialisation Model and Four C's Model, for instance, excel in fostering long-term engagement, while the Traditional Model supports initial integration but may limit personal identity expression. The study also addresses potential drawbacks, such as resource intensity or context-specificity, proposing mitigation strategies like phased implementation and role development.

Ultimately, this research provides actionable insights for organisations aiming to enhance their newcomer integration strategies, offering practical recommendations such as adopting diverse onboarding models and implementing flexibility, focusing on mentorship and feedback, prioritising cultural fit and integration, tailored to diverse organisational needs. The authors suggest further research through surveys with newcomers to validate their findings. Managers may struggle to move away from the old traditional practices and fully support newcomers' personal strengths. Provide appropriate training for managers to help them understand and implement the principles of this model effectively.

Explainable Artificial Intelligence and Social Theory Integration for Advancing Educational Equity in Nepal

2025-11-10T08:49:05+00:00

This chapter examines entrenched socioeconomic disparities in Nepal’s education system through the integration of explainable artificial intelligence (XAI) and foundational social theories of equity. While Nepal has made progress in enrollment, persistent gaps in access, retention, and learning outcomes remain among groups marginalized by caste, gender, and geography. Existing policy analyses often rely on linear statistics or descriptive methods and lack operational links to sociological theory. To address this lacuna, we develop a mixed-methods framework that blends predictive machine learning with interpretability (SHAP) and qualitative inquiry to ground algorithmic findings in lived experience. Using national-level datasets — notably the Education Management Information System (EMIS) and the Nepal Living Standards Survey (NLSS)—we operationalize a Capability Index and train ensemble models (Random Forest and XGBoost) to predict capability deprivation and dropout risk. SHapley Additive exPlanations (SHAP) are applied to attribute model outputs to observable socioeconomic and school-level features. We formalize the predictive problem and its interpretability as follows: given feature set X = {x₁, . . . , x_n} and an outcome Y (capability index or dropout probability), we estimate \(\hat{Y}\) = f(X; θ) and decompose \(\hat{Y}\) additively into baseline and feature contributions \(\hat{Y}\) = ϕ0 +\(\Sigma\)_i ϕi. This decomposition informs policy levers by quantifying marginal contributions of poverty, distance to school, caste status, and school resources. Beyond technical contributions, the chapter situates model outputs within Sen’s Capability Approach and Bourdieu’s Cultural Capital Theory to interpret how structural constraints and cultural resources shape educational opportunity. Deliverables include a resource allocation framework, SHAP-driven simulation dashboards for policymaking, and early-warning indicators for dropout prevention. Qualitative interviews with educators and community stakeholders are used to validate and contextualize the quantitative results. Together, these elements advance both theory and practice: they demonstrate how XAI can produce socially meaningful, policy-ready evidence for more equitable education in Nepal and similar low- and middle-income contexts.

A Case Study on Exploring YOU Artificial Intelligence and GeoGebra in Learning the Method of Variation of Parameters in Second-Order Differential Equations

2025-11-10T08:58:09+00:00

The study presents an innovative experience carried out with engineering students at the University of Antofagasta, Chile, during the first semester of 2023.

The main objective was to analyse how students describe their learning of the method of variation of parameters when solving a certain second-order linear differential equation, using a combination of the YOU artificial intelligence tool and GeoGebra software.

The research was conducted using a qualitative case study approach, which allowed for an exploration of the perceptions of three students as they used these digital tools. The results show that the integration of both technologies fostered reflection and autonomy in learning. The students evaluated the experience positively, emphasising that these tools do not replace reasoning, but rather enhance it.

In conclusion, the constructivist nature of this type of activity is highlighted, where knowledge is actively built through the interaction between the student and digital technologies. Additionally, YOU supported symbolic reasoning and computation, while GeoGebra provided a visual understanding of the process, turning the resolution of differential equations into a collaborative and participatory experience.

Enhanced Deep Learning Model for Accurate and Automated Detection of Hepatic Steatosis

2025-11-10T09:04:48+00:00

Background: Hepatic Steatosis is one of the most prevalent liver disorders globally. Ultrasound imaging is widely used as the primary screening tool for Hepatic Steatosis. However, its diagnostic performance can vary significantly depending on the operator’s skill and the quality of the equipment. Recent advances in deep learning have brought new opportunities to medical imaging, providing automated, consistent, and quantitative assessments that reduce dependency on operator expertise.

Objectives: This study aims to develop a deep learning (DL)-based framework that enhances the detection and grading of Hepatic Steatosis from ultrasound images. The key goal is to achieve accuracy levels comparable to experienced radiologists while maintaining interpretability and efficiency for real-time use in clinical practice.

Methods: B-mode ultrasound images and cine clips were collected from patients, covering multiple liver views to capture diverse anatomical perspectives. Alongside imaging data, patient metadata such as age, body mass index (BMI), and comorbid conditions were also recorded to enrich the dataset. The proposed system employs a multi-view ultrasound preprocessing approach, followed by transfer learning to leverage existing feature representations. Attention-driven convolutional neural networks (CNNs) are then used to capture fine details across image regions. To ensure clinical usability, explainability modules are integrated, allowing transparent interpretation of model predictions.

Findings: Experimental evaluation demonstrated that the framework outperformed traditional single-view methods, offering improved sensitivity and specificity in detecting hepatic Steatosis. The performance was closely aligned with radiologist-level assessments. Furthermore, the system showed low latency, highlighting its suitability for near-real-time diagnostic applications.

Conclusion: Unlike conventional models that rely on a single static image, this study introduces a multi-view fusion strategy enhanced with attention mechanisms and explainability tools. This combination not only strengthens predictive accuracy but also ensures transparency and trustworthiness, critical factors for adoption in clinical settings. Despite the promising performance, challenges such as data variability, subtle early-stage disease patterns, and model interpretability remain. Addressing these limitations through larger, diverse datasets and explainable AI approaches will be essential for translating these models into clinical practice.

Classification of Associative Algebras Satisfying Quadratic Polynomial Identities

2025-11-10T09:07:15+00:00

In quantum mechanics, associative algebras play an important role in understanding symmetries and operator algebras, providing new algebraic frameworks for describing physical systems. This work classifies associative algebras over a field K that are generated by a finite set G and satisfy a polynomial identity of the form X² = aX + b, where a and b are elements of K and X varies either over all elements of the algebra or over all elements of the multiplicative semigroup S generated by G. One of the results obtained in this work shows that algebras satisfying X² = 0 over fields of characteristics different from 2 are nilpotent of index 3.

The results obtained were validated computationally using the GAP system.

Analysing Image Processing Algorithms Using Correlational Values within the Cloud Platform

2025-12-31T06:27:31+00:00

Background: Image processing strategy is an important part of image processing to visualise the performance and outcome of the goal. Image processing is a discipline in which the process's input and output are both images. It is a process that entails elementary operations such as noise reduction, contrast enhancement, and image sharpening. Image analysis is a process that takes images as inputs but produces attributes extracted from those images as outputs (e.g., edges, contours, and the identity of individual objects).

Aims: This paper aims to analyse the algorithms of image processing in the cloud platform. Several algorithms are commonly used in image processing and computing techniques. Correlations for the observation matrix were observed to marginalise the images, and results were transmitted to the cloud platform.

Methodology: Here, a selection of state-of-art is applied to test image processing execution and timing factor using different strategies and platforms. Among them, the dataset structure and performance of the system can choose a verification algorithm to achieve the final operation. Based on the structure of a real-time image processing system based on SOPC technology is built, and the corresponding functional receiving unit is designed for real-time image storage, editing, viewing, and analysis. Datasets were collected online from the free domain of kaggle.com. Images belong to the traffic light of 250 out of 2056 files. 120 images were selected randomly to process after pre-processing of the images.

Results: Studies have shown that the image processing system based on cloud computing has increased the speed of image data processing by 12.7%. Compared with another platform, especially in the case of segmentation and enhancement of the image. This analysis has advantages in image compression and image restoration on a cloud platform. Qualitative and quantitative performances in the cloud platform of the algorithm are compared, and the results of the three indicators show that the platform has better performance than others. The results show that the cloud platform requires less computational time in comparison with others after loading the image file into the system.

Conclusion: Different image processing parameters like noise, smoothing, the timing of enhancement and segmentation have a greater effect on the compression effect of the image, including correlational value within the dataset of the image. The larger the correlation, the less compressed the image data is, the faster the image compression rate, and the lower the image's peak entry-to-noise ratio.

Fractional Pharmacokinetic Models in One-compartment Systems

2025-12-31T06:30:32+00:00

Classical models typically use integer-order (ordinary) differential equations and predict exponential-like concentration decay after administration of drugs. These classical assumptions are adequate for many compounds but fail to capture anomalous behaviours seen in numerous drugs: long tails in concentration-time curves, non-exponential elimination, or irregular accumulation after repeated dosing. The study aimed to develop fractional pharmacokinetic models in one-compartment systems to enhance drug absorption. Fractional derivatives can be inserted into compartmental networks to create fractional multi-compartment models. Recent theoretical work provides a general framework for embedding fractional orders within compartmental mass-balance systems while preserving physically meaningful constraints (mass conservation and positivity), which is important for physiological interpretability. Practical implementations often use efficient strategies to reduce the cost of history terms (e.g., short-memory approximations, nonuniform time grids, or convolution quadrature approaches). Fractional models can inform controlled-release formulation design and the prediction of long-term toxicological accumulation. However, adoption in clinical pharmacology requires standardised parameter-estimation pipelines, software, and regulatory acceptance—areas currently under development. Widespread adoption will require advances in parameter estimation, computational tools, and translational validation, but the literature over the past two decades demonstrates clear progress and growing interest in fractional approaches.

Hardware Implementation and Optimisation of a Chaotic-Chua-Based Pseudo-Random Number Generator for Security Applications

2025-12-31T06:32:34+00:00

In this paper, we propose a high-performance pseudo-random number generator (PRNG) based on the Chua chaotic system, specifically optimised for FPGA deployment. The inherent nonlinear dynamics and extreme sensitivity to initial conditions exhibited by the Chua system substantially improve the statistical quality of the generated sequences, yielding high entropy and strong resistance to cryptanalytic attacks. A hardware-oriented architecture is carefully designed to minimise arithmetic complexity, enhance parallelism, and ensure efficient utilisation of FPGA resources. The proposed implementation achieves a maximum operating frequency of 104.613 MHz and an ultra-high throughput of 3,347.616 Mb/s, making it well-suited for real-time and lightweight security applications. Experimental results confirm that the proposed Chua-based PRNG delivers excellent performance in terms of unpredictability, statistical robustness, and hardware efficiency.

Ethical Artificial Intelligence in Education: A Southern African Input–Process–Output (IPO) Governance Framework

2026-01-14T10:13:44+00:00

Artificial Intelligence (AI) holds colossal promise for advancing human development, particularly in education, science, and communication. The ethical issues surrounding the design and use of artificial intelligence (AI) have become more important as it becomes more common in schools, government, and society as a whole. The Input–Process–Output (IPO) Ethics Framework is a complete paradigm for finding and dealing with ethical hazards at every stage of the AI lifecycle in response to these problems. This chapter explores the social and ethical considerations of artificial intelligence (AI) as it integrates into education and society. It examines challenges such as data privacy, algorithmic bias, AI trustworthiness, and human agency. The literature highlights context, human agency, and the importance of diverse stakeholder involvement in AI governance, AI literacy, responsible education, and strategies for ethical assessment and mitigation. A literature review of recent articles and policy documents informs this study, focusing on AI’s evolving role in education. Education for AI focuses on Training AI Experts, preparing the Workforce and Public AI Literacy. AI for Education leverages AI tools to enhance teaching, learning, and administrative processes in educational systems. The research develops an input-process-output (IPO) framework to address ethical concerns at each stage of AI development. The IPO model outlines the ethical implications for the input, process, and output phases. Section one addresses AI’s social implications. This chapter also examines AI educational policy using the United Nations Educational, Scientific and Cultural Organisation’s (UNESCO’s) guidelines as a benchmark for member states. Ethical considerations in AI development and usage were also discussed in this chapter. Finally, this chapter presents the AI IPO Ethical Framework, detailing ethical responsibilities at each stage. The study underscores the role of policymakers, researchers, and higher education institutions in shaping AI’s ethical trajectory. It emphasises responsible AI implementation, ensuring that AI systems are developed and deployed with ethical considerations in mind. The proposed framework serves as a guiding tool for assessing ethical risks and ensuring responsible AI integration in education. By fostering AI literacy and ethical awareness, this study contributes to ongoing discussions on AI ethics, advocating for transparent, fair, and accountable AI practices. It aims to support the ethical advancement of AI in education and governance.

AI-Driven Wafer Inspection: Deep Learning, Transformers, and Generative Models for Defect Analysis in Semiconductor Manufacturing

2026-01-14T10:20:19+00:00

Semiconductor manufacturing at advanced technology nodes demands inspection systems capable of identifying increasingly subtle, stochastic, and mixed-type defects. Traditional rule-based and handcrafted-feature approaches are no longer sufficient to address the complexity of modern wafer patterns, prompting the integration of artificial intelligence (AI) into high-volume manufacturing workflows. This chapter presents a unified framework for AI-driven wafer inspection that combines convolutional neural networks (CNNs), Vision Transformers (ViTs), and generative models such as variational autoencoders (VAEs) and generative adversarial networks (GANs). CNNs are effective for wafer map pattern classification, learning hierarchical spatial features that capture centre, edge-ring, and composite wafer failure modes. Transformers extend this capability by modelling long-range spatial dependencies, enabling improved performance in optical and SEM imaging scenarios where global context is essential. Generative models enhance sensitivity to rare or previously unseen defects by learning the underlying distribution of defect-free patterns and detecting deviations through reconstruction error or anomaly scoring.

A hybrid ViT–GAN architecture is introduced to demonstrate how discriminative and generative pathways can be fused to deliver high accuracy and low false-alarm rates across diverse defect classes. Extensive comparisons using public datasets and synthetic SEM-like datasets show that AI models substantially outperform classical techniques, particularly in low-sample regimes and in the presence of noise, illumination variations, or rotation. Deployment considerations, including inference speed, model compression, domain adaptation, and explainability, are discussed to highlight practical challenges in integrating AI into semiconductor fabs. The chapter concludes with emerging trends such as self-supervised learning, large vision models, multimodal data fusion, and temporal defect modelling, which are expected to shape the next generation of intelligent wafer inspection systems.