Hung Yen University of Technology and Education

In the past decade, numerous deep learning-based object detection approaches have been released and obtained desirable performance when executing in favorable weather conditions. Nevertheless, these detectors attain incomplete results of detecting objects in the rain, a frequently occurring weather condition, due to a drop in visibility and lack of crucial features for depicting objects. In this work, to bridge this gap, we publish a novel approach for elevating the efficiency of object classification and localization impaired by rain, termed UFA-Net. Our proposed approach achieves attractive object detection results via absorption sharp features at diverse scales generated from rainy images. To this end, the UFA-Net consists of three subnetworks, namely, a feature restitution (FR) subnetwork, an unsupervised learning (UL) subnetwork, and an object detection (OD) subnetwork, where, the FR subnetwork provides robust information to the OD subnetwork through the UL subnetwork. In our architecture, the FR and UL subnetworks are solely activated during training progress, and the OD subnetwork is accountable for estimating the object localization and its label in rainy weather conditions. The exhaustive experimental results substantiate that our UFA-Net achieves the highest mean average precision (mAP) scores on published light, medium, heavy, synthetic, and natural rainy image sets, respectively, surpassing compared object detection models and combinations of detectors and rain removal methods.

This paper investigates the finite-time dissipativity problem for uncertain tempered fractional-order neural networks with constant time delays and external disturbances. By employing a novel analytical framework combining the inf-sup technique, Laplace transform for tempered Caputo derivatives, generalized Grönwall inequality, and linear matrix inequality (LMI) conditions, new tractable and less conservative finite-time dissipativity criteria are derived. The proposed approach ensures both transient state boundedness and input-output energy dissipation performance over finite time horizons, offering stronger performance guarantees than classical finite-time stability. Numerical simulations are presented to validate the theoretical results.

Vietnam’s wooden forest products industry is an important export sector, contributing to industrial growth and employment. However, it is facing increasing pressures related to challenges such as forest and export sustainability. Despite its potential, Vietnam’s export performance remains uneven across destination markets, related to the presence of significant unrealized trade potential. This study examines the determinants of export efficiency in Vietnam’s wooden forest products sector by moving beyond traditional gravity variables to incorporate institutional and cultural dimensions. Using a panel of 70 trading partners between 2004 and 2023, covering more than 93% of Vietnam’s total wood exports, this study employs an instrumental-variable single-stage stochastic frontier gravity model (IV-SFGM) to estimate trade potential. The results show that economic size, favorable exchange rates, and shared borders significantly enhance export performance. Furthermore, geographical distance and land enclosure remain persistent structural barriers, particularly relevant for bulky and logistics-intensive wood products. Institutional and cultural distance constitute substantial non-tariff barriers, significantly reducing export efficiency across markets. Conversely, regional trade agreements, trade freedom, and foreign direct investment play a critical role in mitigating inefficiencies and facilitating market penetration. Export efficiency in Vietnam’s wooden forest products sector indicates considerable improvement, rising from approximately 25% in the mid-2000s to over 55% in recent years, indicating notable progress in the market and highlighting considerable untapped potential. So, integrating institutional and cultural factors into a frontier-based gravity framework, this study offers novel empirical evidence from an emerging, biodiversity-rich economy with evolving governance institutions. The findings provide important policy implications for aligning export growth with institutional reform and trade liberalization, thereby contributing to the achievement of SDGs such as Decent Work and Economic Growth.

Independent control of the coercivity ( H-C ) and exchange bias ( H-EB ) in ferromagnetic/antiferromagnetic ( FM / AFM ) heterostructures is essential for optimizing perpendicular spintronic devices, including magnetic random access memory (MRAM) and nanoscale magnetic sensors, yet it remains challenging due to their shared interfacial origin. In this work, we demonstrate a systematic approach to decouple these properties in [Co/Pd] multilayers (MLs) exchange-coupled to FeMn by engineering the stacking sequence and antiferromagnetic layer thickness ( t(FeMn) ). Two configurations-top-pinned [Co/Pd]/FeMn and bottom-pinned FeMn/[Co/Pd] -were fabricated using magnetron sputtering. Despite exhibiting a similar f(cc)(111) texture, the two structures display markedly different magnetic responses. The bottom-pinned configuration promotes rotatable uncompensated spins, resulting in a significant enhancement of H-C (up to 1440 Oe) with negligible H-EB . On the contrary, the top-pinned configuration stabilizes pinned uncompensated spins, leading to a pronounced perpendicular exchange bias (EBs) (up to 188 Oe) with a minimal impact on H-C . By varying t(FeMn)H(C) , and H-EB can be tuned independently over wide ranges. These results demonstrate that the stacking-sequence-dependent interfacial spin configurations govern magnetic behavior rather than bulk structural differences, providing a practical and scalable strategy for designing high-stability reference layers and low-switching-energy free layers in advanced spintronic devices.

This paper presents a systematic comparative study of bio-inspired optimization algorithms for tuning Proportional–Integral–Derivative (PID) controllers in brushless DC (BLDC) motor speed regulation. Four representative metaheuristics are considered, namely the Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Grey Wolf Optimizer (GWO), and Butterfly Optimization Algorithm (BOA). Unlike many previous works that focus on a single algorithm or on limited pairwise comparisons, this study develops a unified optimization and simulation framework in which all methods are applied to the same BLDC motor model, under identical operating conditions and performance criteria. The PID gains are obtained by minimizing the Integral of Time-weighted Absolute Error (ITAE), while controller performance is independently assessed using the Integral of Absolute Error (IAE), Root Mean Squared Error (RMSE), and key transient response metrics to rigorously quantify tracking accuracy, dynamic behavior, and disturbance rejection. Numerical results clearly show that all bio-inspired approaches outperform conventional fixed-gain PID control. More importantly, the comparative analysis highlights the specific strengths of each algorithm: PSO provides the most balanced and robust performance over all indices, GWO achieves very fast response with low overshoot, and BOA yields an almost overshoot-free response at the expense of slower convergence. These findings not only confirm the effectiveness of bio-inspired tuning, but also offer practical guidelines for selecting appropriate optimization strategies in high-performance BLDC motor drive applications.