Authors: Dr. Daniel Foster, Dr. Olivia Bennett, Ethan Clarke, Dr. Hannah Mitchell, Andrew Richard

Abstract: The rapid growth of deep learning has enabled state-of-the-art performance across vision, speech, and natural language processing tasks, driving widespread adoption in both academic research and industrial applications. However, this progress has been accompanied by a steady increase in model depth, parameter count, and computational complexity, which poses significant challenges for deployment in resource-constrained environments such as mobile devices, embedded systems, and edge computing platforms with limited memory, power, and latency budgets. To address these constraints, this article presents a comprehensive review of model compression and knowledge distillation techniques developed between 2000 and 2021, synthesizing foundational methods including network pruning, low-precision quantization, and entropy-based coding, as well as teacher–student learning paradigms that transfer representational and decision-level knowledge from large, overparameterized models to compact alternatives. Using representative architectural and training diagrams, we illustrate how these approaches systematically reduce memory footprint and computational cost while preserving, and in some cases improving, predictive accuracy. Finally, we examine key empirical findings across vision, speech, and language domains, identify persistent limitations related to generalization, hardware efficiency, and evaluation methodology, and outline future research directions toward scalable, energy-efficient, and deployable AI systems.

DOI: https://doi.org/10.5281/zenodo.20052481

Published by: vikaspatanker

Authors: Dr. Alexander Hayes, Dr. Natalie Brooks, Ryan Cooper, Dr. Victoria Simmons, Andrew Richard

Abstract: The rapid growth of deep neural networks (DNNs) in both model size and deployment scale has placed renewed emphasis on optimization techniques that balance convergence speed, numerical stability, computational efficiency, and resource utilization, particularly as training workloads increasingly span heterogeneous hardware platforms and distributed computing environments. This article presents a systematic analysis of optimization methods for large-scale deep learning, encompassing stochastic first-order approaches such as momentum-based gradient descent, adaptive optimizers that adjust learning rates based on gradient statistics, normalization strategies that stabilize internal representations and smooth optimization landscapes, curvature-aware methods that incorporate second-order information, and system-level techniques including large-batch, mixed-precision, and distributed training. Drawing on publicly available empirical evidence and widely cited foundational studies, we examine how optimization choices shape training dynamics, scalability characteristics, convergence behavior, and final model performance across diverse deep learning workloads. Using representative figures from prior work—including Batch Normalization training dynamics and adaptive optimization formulations—we synthesize practical guidance for selecting and tuning optimization strategies at scale while identifying persistent challenges related to generalization, communication efficiency, and the alignment of optimization algorithms with modern hardware and system architectures.

DOI: https://doi.org/10.5281/zenodo.20052352

Published by: vikaspatanker

Authors: Dr. Daniel Thompson, Dr. Olivia Bennett, James Walker, Dr. Hannah Collins, Andrew Richard

Abstract: Modern distributed systems operate at a scale and complexity that far exceed the limits of manual management and static fault-handling mechanisms, as they span geographically dispersed resources, heterogeneous hardware and software stacks, and dynamically changing workloads. In such environments, failures are not exceptional events but an inherent characteristic of normal operation, arising from partial outages, transient faults, software defects, and unpredictable interactions among system components. Autonomic computing emerged in the early 2000s as a response to these challenges, proposing self-managing systems capable of self-configuration, self-optimization, self-protection, and self-healing through continuous feedback and adaptation. Over the past two decades, advances in artificial intelligence and machine learning have substantially strengthened autonomic control loops, transforming them from rule-driven mechanisms into adaptive, data-driven decision systems that can learn from experience, generalize across failure scenarios, and operate effectively under uncertainty. This article presents a comprehensive overview of AI-driven autonomic control for self-healing distributed systems by synthesizing foundational autonomic computing architectures, closed-loop control models, and learning-based decision mechanisms. Leveraging established architectural diagrams from pre-2021 literature, we analyze how reinforcement learning, probabilistic reasoning, and hybrid AI techniques enhance fault detection, root-cause analysis, and recovery planning, and we conclude by highlighting key empirical studies and open research challenges that continue to motivate advances in intelligent, self-healing distributed systems.

DOI: https://doi.org/10.5281/zenodo.20064010

Published by: vikaspatanker

Authors: Dr. Jonathan Miller, Dr. Emily Carter, Michael Anderson, Dr. Sophia Reynolds, Andrew Richard

Abstract: Autonomous and intelligent systems are increasingly deployed in complex, real-world environments characterized by stochastic dynamics, partial observability, delayed feedback, and continual change, where classical model-based control strategies often struggle due to their reliance on accurate system identification, fixed assumptions, and limited scalability. In response to these challenges, Reinforcement Learning (RL) has emerged as a compelling control paradigm that enables agents to autonomously learn optimal or near-optimal control policies directly through interaction with their environment, leveraging reward-driven feedback rather than explicit system models. This article surveys and synthesizes reinforcement learning-based control mechanisms with a particular emphasis on actor-critic architectures and deep reinforcement learning approaches for continuous control, which have proven especially effective in high-dimensional and nonlinear domains. Drawing on foundational and influential studies published between 2000 and 2021, the discussion examines how RL frameworks facilitate adaptive decision-making, online policy improvement, and robust control under uncertainty, while also addressing critical issues related to convergence, stability, safety, and sample efficiency. Representative applications in robotics, autonomous navigation, and intelligent cyber-physical systems are highlighted to demonstrate practical impact, and publicly available architectural diagrams are integrated to clearly illustrate core learning loops, policy-value interactions, and control workflows, providing a cohesive and accessible reference for researchers and practitioners designing next-generation intelligent autonomous controllers.

DOI: https://doi.org/10.5281/zenodo.20052068

Published by: vikaspatanker

Authors: Varun Garg, Shruti Rajak, Arpita Maravi, Anupama Awadhiya, Sarah Khan

Abstract: The increasing presence of generative artificial intelligence (AI) in educational settings is transforming the way students engage with learning. Tools powered by AI are making information more accessible, enabling quicker completion of academic tasks, and offering personalized support tailored to individual needs. While these benefits are undeniable, there is a growing concern that continuous dependence on such technologies may gradually reduce students’ active cognitive involvement in the learning process. This study explores how the use of generative AI tools influences student learning behaviour, particularly focusing on critical thinking and problem-solving skills. To gain a comprehensive understanding, a mixed-method approach was adopted, combining survey responses with a comparative evaluation of tasks completed with and without AI assistance. The results suggest that although AI enhances efficiency and convenience, overreliance on these tools can limit deeper cognitive engagement and independent reasoning. The findings emphasize the importance of mindful and balanced use of generative AI in education, ensuring that technological support complements rather than replaces essential learning processes.

DOI: https://doi.org/10.5281/zenodo.20050504

Published by: vikaspatanker

Authors: Dr. Prahlad Singh

Abstract: Mathematics was often portrayed as the story of great men, but history proves otherwise. In ancient times, the Early Modern Era, during the development of the research university in the nineteenth century, and through the modern scientific state, women played roles in the development of mathematical thought and practice. They commented on mathematics, taught mathematics, produced textbooks, and made groundbreaking contributions to number theory, algebra, logic, geometry, elasticity theory, and computing. However, their success did not always translate into recognition. Women lacked access to education, entry into academies and universities, resorted to anonymous publications, received precarious or unwaged appointments, and were known for their relationships with eminent men. In this essay, it will be argued that the major historical trend lies in women's active involvement in mathematics coupled with their systematic exclusion from the processes of certifying achievements, awarding rewards, archiving accomplishments, and recalling mathematicians in institutional memory. Among the most notable women who worked in the field of mathematics will be mentioned Hypatia, Maria Gaetana Agnesi, Sophie Germain, Sofia Kovalevskaya, Emmy Noether, Grace Hopper, Julia Robinson, Euphemia Lofiton Haynes, Maryam Mirzakhani, Karen Uhlenbeck, and Maryna Viazovska.

DOI: https://doi.org/10.5281/zenodo.20049877

Published by: vikaspatanker

Authors: Dr. Prahlad Singh

Abstract: Ancient Indian mathematics and astronomy evolved as part of a tightly linked scholarly tradition wherein numbers, geometry, trigonometry, calendars, and planetary astronomy were practiced and theorized simultaneously rather than as distinct and separate activities. In such an intellectual environment, Aryabhata, whose Aryabhatiya was written around 499 CE, emerged as the pivotal classical scholar. He provided not only mathematical formulas but also geometrical facts, sine values, solutions to indeterminate equations, models for planetary motion, theories of eclipses, and the remarkable proposition that the observed daily rotation of the stars is due to the rotation of the Earth. But what made Aryabhata influential was more than just his own contributions, since other Indian mathematicians and astronomers such as Bhaskara I, Brahmagupta, Lalla, and Bhaskara II preserved, disputed, modified, and elaborated on his teachings, resulting in one of the most impressive pre-modern mathematical astronomical traditions found anywhere in the world. This paper contends that the accomplishments of Aryabhata and the Indian tradition based on his work can be considered an important scholarly tradition in which mathematics worked for astronomy, astronomy inspired mathematics, and the practice of commentary was a powerful force of innovation.

DOI: https://doi.org/10.5281/zenodo.20049527

Published by: vikaspatanker

Authors: Pradeep Kumar, Dr. Sunil Maggu

Abstract: Brain tumors represent life-threatening neurological conditions requiring precise classification for effective treatment planning. This paper presents a Multi-Class Brain Tumor Classifier capable of distinguishing between Glioma, Meningioma, Pituitary, and No Tumor classes from MRI scans. Unlike standard binary classifiers, the system employs an Ensemble of five supervised Machine Learning algorithms — Random Forest, XGBoost, SVM, KNN, and Naive Bayes — combined through Soft Voting for robust decision-making. Texture Analysis using GLCM (Gray-Level Co-occurrence Matrix) and LBP (Local Binary Pattern) feature extraction provides explainable, biologically interpretable features rather than opaque deep-learning representations. The system is deployed as a Flask web application that automatically generates standardized PDF Medical Reports for clinical documentation. Experimental evaluation on the Kaggle Brain Tumor MRI Dataset (7,023 images) confirms that the ensemble approach achieves superior accuracy, with Random Forest and XGBoost leading individual classifier performance at 90.68% and 90.53% respectively.

Published by: vikaspatanker

Authors: Prof.F.J.Sayyad, Kale Tejas Popat, Ganeshkar Shraddha Santosh, Kucheker Priti Dattaray

Abstract: Electric vehicles (EVs) represent a promising and sustainable mode of transportation that reduces greenhouse gas emissions and dependence on fossil fuels. battery and wiring harness playing key roles. This abstract provides an overview of the selection of batteries and wiring harnesses for electric vehicles. Battery selection involves evaluating various parameters, including energy density, power density, cycle life, and cost. Lithium-ion batteries are the most commonly used technology due to their high energy density, long cycle life, and low self-discharge rates. The wiring harness in an electric vehicle is a complex network of wires and connectors that connects various electrical components, including the battery, motor, inverters, and other vehicle systems appropriate wiring harness is critical to ensure the efficient flow of power and data throughout the vehicle.

Published by: vikaspatanker

Authors: Paul Binu, Paulu Wilson, Ronal Shoey George

Abstract: This paper presents an emotion-aware voice-based conversational therapy assistant that integrates speech recognition, con-versational AI, and emotional text-to-speech synthesis into a unified pipeline. The system captures user speech through a microphone, transcribes it to text, generates context-aware empathetic responses using a large language model (Gemini AI), and synthesizes emotion-ally expressive speech output using IndexTTS2 with zero-shot voice cloning. The architecture follows a modular design comprising four major modules: Voice Input, Processing and AI, Emotion Analysis, and Speech Synthesis. The emotion mapping subsystem identifies user affect and selects an appropriate response emotion to guide TTS output. Evaluation against two baselines (generic neutral TTS and rule-based keyword approach) demonstrates that the proposed model achieves the highest overall score of 74.51, significantly outper-forming both baselines in holistic end-to-end quality. The system balances emotion recognition accuracy, response relevance, and audio naturalness, making it suitable for mental health support, virtual assistants, and human-centered AI applications. The results confirm that combining emotional conditioning with contextual response generation yields substantially better conversational quality than neutral or rule-driven approaches.

DOI: https://doi.org/10.5281/zenodo.20045982

Published by: vikaspatanker