2026-03-17
AMINA KESSENTINI et al.
The escalating demand for high-resolution multimedia content has necessitated more efficient video compression solutions. The Versatile Video Coding (VVC) standard, despite achieving remarkable compression gains, introduces significant computational complexity, primarily due to its exhaustive Rate-Distortion Optimization (RDO) process. To address this, we propose an intelligent approach leveraging supervised machine learning techniques to streamline the VVC encoding process. Specifically, we introduce a Lightweight Neural Network (LNN) for efficient coding unit partitioning decisions and a Decision Tree (DT) classifier for optimizing the intra prediction process. This dual-method framework, tailored for All Intra coding configuration, significantly reduces encoder complexity while maintaining compression performance and visual quality. Through extensive testing, we demonstrate a remarkable 65.47\% reduction in encoding time with minimal impact on compression efficiency and no perceptible degradation in video quality. These findings represent a significant step towards making high-efficiency VVC encoding more practical for real-world applications.
2026-03-17
KOUSAR SALEEM et al.
Genomic data sharing has become an essential component of biomedical research, enabling large-scale collaborations and accelerating discoveries in human genetics. To balance the need for accessibility with privacy concerns, several controlled-access mechanisms have been proposed, including genomic beacons. Genomic beacons answer simple presence/absence queries about specific genetic variants. However, prior work has demonstrated that beacons remain vulnerable to genome reconstruction attacks, where an adversary can recover large portions of participants’ genomes using summary statistics. Building on insights from prior reconstruction attacks, we introduce an approach that unifies SNP correlation and allele frequency alignment objectives within a single-stage joint optimization framework. Unlike earlier two-stage methods that alternately minimized correlation and frequency losses, our joint formulation improves both reconstruction accuracy and computational effeciency, achieving an average F1-score of 71.4%, a 1.4 percentage point improvement over the state-of-the-art method and substantially outperforming the baseline approach (45%), while reducing the runtime for reconstructing 2000 SNPs across 100 individuals from 10 to 7.4 h, marking a 26% decrease in computational cost. These results underscore the pressing need for more robust privacy-preserving mechanisms in genomic beacon protocols.
2026-03-17
DUSHMANTA KUMAR DAS et al.
Maintaining smart grid stability is crucial for the reliable operation of decentralized electricity networks, especially as the energy sector becomes more complex. The process of ensuring grid stability begins with collecting consumer data and comparing it to power supply requirements. Ultimately, consumers receive a report showing their energy use and pricing details. However, this process is time-consuming and can be improved by leveraging artificial intelligence to predict smart grid stability more efficiently. Specifically, an optimized Long Short-Term Memory (LSTM) network is proposed to predict smart grid stability, addressing the challenges associated with traditional data collection and evaluation methods. Simulations from previous studies on a decentralized smart grid control (DSGC) system, modeled in a star topology with four nodes and tested under both stable and unstable grid conditions, are used to support the development of the proposed model. The performance of the proposed model was then compared with existing algorithms such as PSO-based LSTM, CTO-based LSTM, and LSTM classifiers. A confusion matrix was calculated for each to assess their effectiveness. The SGSC-KKO-LSTM classifier demonstrated strong performance, achieving an accuracy of 99.93%, a recall of 99.97%, a specificity of 99.91%, a precision of 99.85%, an F1-score of 99.90%, and a minimal misclassification rate of just 0.01%. These findings highlight the effectiveness of LSTM in enhancing forecasting accuracy and operational efficiency, offering valuable insights into grid performance under various conditions.
2026-03-17
SAMET TENEKECİ et al.
Software Size Measurement (SSM) is crucial for estimating required project effort as well as budget and schedule. However, many small and medium-sized companies struggle to apply objective SSM due to limited resources and lack of expertise. This often leads to inaccurate estimates and project overruns. There is a need for practical, low-resource solutions that support these tasks without requiring expert involvement. Motivated by this challenge, this study proposes an automated software size measurement approach that formulates the measurement task as supervised regression over natural language requirements, using domain-adapted transformer models. We construct large-scale Turkish and English software engineering corpora to pre-train two models: SE-BERT and SE-BERTurk. These models are fine-tuned on a multilingual, organization-specific dataset annotated with COSMIC Function Points (CFP) and MicroM size by domain experts. We evaluate the models using various regression and classification metrics. Results show that SE-BERT improves exact match accuracy from 66.9% to 68.2% compared to BERT, while SE-BERTurk improves from 65.7% to 69.3% over BERTurk. Both models also achieve lower normalized errors than previous domain-adapted baselines BERT_SE and RE-BERT, demonstrating superior generalization. These findings highlight the effectiveness of domain-specific pre-training for software engineering tasks and its potential to support accurate software size estimation, especially in low-resource languages like Turkish and in real-world, organization-specific contexts.
2026-01-18
ZÜLAL HİLAL YILDIZ BUDAK et al.
Traffic signal management is a critical challenge due to its environmental, economic, and public health impacts. The maximum weighted flow method (MaxWeightedFlow) was developed to optimize traffic flow at isolated and coordinated urban intersections. This study proposes a new method, the novel MaxWeightedFlow, which includes two key strategies to enhance the classical approach. The first strategy reduces computational burden by estimating vehicle approach times based on instantaneous speeds, improving real-time performance. The second employs regression analysis to optimize the alpha parameter, representing the vehicle waiting coefficient. The proposed approach, the novel MaxWeightedFlow, was evaluated using real-world traffic data from Kilis, Türkiye, and demonstrated significant performance improvements in comparison to the existing benchmarks within the Simulation of Urban Mobility (SUMO) simulation environment. Specifically, for isolated intersections, the method resulted in a 0.45% reduction in average travel time and a 3.74% reduction in average waiting time when compared to the classical MaxWeightedFlow method. Moreover, it led to a 0.24% increase in average speed, indicating enhanced traffic flow efficiency. For coordinated intersections, the novel MaxWeightedFlow reduced the average travel time by 2.63% and the average waiting time by 17.52%. Additionally, the approach improved the average speed by 1.60%, further underscoring its effectiveness in optimizing traffic dynamics. These findings underscore the effectiveness of the new method in enhancing traffic flow efficiency, leading to faster traffic management and higher operational efficiency, which aligns closely with real-world traffic measurements.