Simona-Vasilica Oprea, Adela Bâra^*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.078985 - 09 April 2026

Abstract Quantization has emerged as an important technique for enabling efficient deployment of large language models (LLMs) by reducing their memory and computational requirements. This research conducts an evaluation of INT8 quantization on several state-of-the-art LLMs, GPT-2, LLaMA-2-7B-Chat and Qwen1.5-1.8B-Chat, across two hardware configurations: NVIDIA RTX4070 Laptop GPU and RTX4080 Laptop GPU and two tasks: text and code generation. By comparing quantized INT8 models with their FP16 counterparts and a human-written reference, we quantify the trade-offs between performance and efficiency using standard natural language generation metrics (BLEU, ROUGE-1, ROUGE-L) and semantic analysis via GPT-4o and Gemini… More >

Osamah Mohammed Alyasiri^1,2, Yu-N Cheah^1,*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.078927 - 09 April 2026

Abstract Term weighting scheme and feature selection are two fundamental components in text classification (TC) systems, particularly in high-dimensional, multi-class, and imbalanced settings. Term weighting schemes aim to improve document representation by emphasizing discriminative terms across classes, while feature selection (FS) seeks to reduce dimensionality, eliminate irrelevant and redundant features, and enhance classification efficiency and effectiveness. However, most existing studies focus on FS independently of the term-weighting strategy used during document representation, thereby limiting the potential benefits of their interaction. This study addresses this gap by pursuing two main objectives. First, it employs an enhanced supervised… More >

Waqar Ali¹, May Altulyan², Ghulam Farooque³, Siyuan Li⁴, Jie Shao^4,5,*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.078599 - 09 April 2026

Abstract Federated recommender systems (FedRS) enable collaborative model training while preserving user privacy, yet they remain vulnerable to adversarial attacks, unreliable client updates, and misaligned incentives in decentralized environments. Existing approaches struggle to jointly preserve personalization, robustness, and trust when user data are highly non-IID and recommendation quality is governed by ranking-oriented objectives. To address these challenges, we propose a Trustworthy Federated Recommender System (T-FedRS) that extends federated neural collaborative filtering by integrating a ranking-aware reputation mechanism and a lightweight blockchain layer for transparent incentive allocation. Personalization is preserved through locally maintained user embeddings, while item parameters… More >

Sathish Rao Udupi, Gururaj Bolar, Manjunath Shettar^*, Ashwini Bhat

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.078574 - 09 April 2026

Abstract Glass fiber-reinforced polymer composites (GFRPCs) are extensively utilized in the aerospace, automotive, and structural sectors; nevertheless, their heterogeneous and abrasive characteristics result in rapid tool wear during drilling. Drill flank wear among various wear mechanisms notably influences hole quality and dimensional accuracy. This research investigates the impact of spindle speed, feed rate, and drill diameter on flank wear during dry drilling of GFRPC laminates with high-speed steel (HSS) twist drills. A full-factorial design with 81 experiments is used to create a comprehensive dataset. ANOVA indicates that spindle speed is the dominant factor affecting wear changes,… More >

Nourdine Aliane^*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.078545 - 09 April 2026

Abstract The rapid advancement of unmanned aerial vehicle (UAV) technologies has increased demand for flexible autopilot platforms suitable for both research and education. Among available options, the open-source Pixhawk/PX4 autopilot has emerged as a leading solution due to its modular architecture and robust software ecosystem. This survey examines the adoption of the Pixhawk/PX4 platform in research and education. The survey covers the analysis of the Pixhawk/PX4 autopilot software development APIs, its compatibility with ROS middleware and MATLAB/Simulink environments, and environments for software/hardware-in-the-loop simulations. Additionally, it explores the integration of Cutting-Edge technologies to enhance UAVs performance. By More >

Hongjiang Wang¹, Jian Yu², Tian Zhang³, Na Ren⁴, Nan Zhang², Zhenyu Liu^1,*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.077997 - 09 April 2026

Abstract The rapid deployment of Industrial Internet of Things (IIoT) systems, such as large-scale photovoltaic (PV) power stations in modern power grids, has created a strong demand for edge-intelligent fault localization methods that can operate reliably under strict computational and memory constraints. In this work, we propose an edge-intelligent photovoltaic fault localization framework that integrates intelligent computation with classical sub-pixel optimization. The framework adopts a modular, edge-oriented design in which a radial basis function (RBF) network is first employed as a lightweight screening module to enable conditional execution, thereby reducing unnecessary computation for non-faulty samples. For… More >

Giuseppe Marannano^*, Antonino Cirello, Tommaso Ingrassia

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.077884 - 09 April 2026

Abstract In response to the growing need for adaptive optimization algorithms capable of handling complex, multimodal, and high-dimensional search spaces, this paper introduces the Structured Random Cycle-guided Algorithm (SRCA). SRCA is not presented as a fundamentally new optimization paradigm, but rather as an architectural synthesis and a unified adaptive framework for dynamic operator selection. Based on a cycle-structured architecture, directional and stochastic search behaviors are dynamically selected at the individual level. The algorithm orchestrates well-established structured movements with a diverse pool of stochastic exploration strategies, enabling a coherent and adaptive balance between exploration and exploitation throughout More >

Jiajia Liu^1,2, Shuchen Pang³, Peng Xie³, Haitao Zhou³, Chenxi Du³, Haoran Hu³, Bo Tang³, Jianhua Liu³, Fei Jia¹, Huibing Zhang^1,*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.077824 - 09 April 2026

Abstract In the unmanned aerial vehicle (UAV) assisted edge computing system, the broadcast characteristics of the UAV signal, the high mobility of the UAV, and the limited airborne energy make the task offloading strategy face challenges such as increased risk of information disclosure, limited computing resources, and the trade-off between energy consumption and flight time. To address these issues, we propose a K-means in-depth reinforcement learning algorithm based on Soft Actor-Critic (SAC). The proposed method first leverages the K-means clustering algorithm to determine the optimal deployment of ground jammers based on the final distribution of mobile… More >

Sihan Wang¹, Luyao Liu², Xingjun Wang^3,*, Yifan Liu^3,*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.077791 - 09 April 2026

Abstract This paper presents a novel dual-stage approach for efficient gait phase recognition and trajectory prediction, tailored for the operation of wearable devices such as exoskeletons. By leveraging dynamic template matching techniques and addressing their computational challenges, we propose an innovative algorithm that significantly enhances both prediction accuracy and computational efficiency. The approach integrates Dynamic Time Warping-KMeans (DTW-KM) template selection in the offline phase and a Soft Constraint Weighted (SCW) template matching technique in the online phase. In the offline stage, the DTW-KM method extracts diverse and generalizable gait patterns from a database, establishing a robust More >

Seohyun Yoo, Joonseo Hyeon, Jaehyuk Cho^*

CMC-Computers, Materials & Continua, Vol.87, No.3, 2026, DOI:10.32604/cmc.2026.077579 - 09 April 2026

Abstract Despite remarkable advances in medical large language models (LLMs), their deployment in real clinical settings remains impractical due to prohibitive computational requirements and privacy regulations that restrict cloud-based solutions. Small LLMs (sLLMs) offer a promising alternative for on-premise deployment, yet they require domain-specific fine-tuning that still exceeds the hardware capacity of most healthcare institutions. Furthermore, the impact of multilingual data composition on medical sLLM performance remains poorly understood. We present a resource-efficient fine-tuning pipeline that integrates Quantized Low-Rank Adaptation (QLoRA), Fully Sharded Data Parallelism (FSDP), and Sequence Packing, validated across two model scales: MedGemma 4B… More >