This study aims to investigate the finite-time control problem for a class of strict-feedback time-delay nonlinear systems with unknown functions. The control design is based on a fast finite-time practical stability criterion. Unknown nonlinear functions can be estimated using the universal approximation performance of neural networks. Finite-time control design is performed using adaptive backstepping technology. By performing closed-loop stability analyses and choosing appropriate Lyapunov–Krasovskii functionals, all signals in a closed-loop system can be bounded within a finite time. Subsequently, the proposed control method can be applied for the excitation control of synchronous generators. The effectiveness of the proposed method is verified using a numerical model of a single-machine power system.

Unilateral spatial neglect (USN) is a disorder characterized by the inability to attend to the space opposite the cerebral hemisphere lesion, significantly hindering daily activities. Due to the complex neural circuitry of the brain, understanding the mechanisms of USN has proven challenging. In clinical settings, the Behavioral Inattention Test (BIT), a paper-based examination, is commonly used to assess USN. However, improved scores on this test do not necessarily guarantee functional improvement, as it solely evaluates performance in a two-dimensional space. To address this limitation, various approaches utilizing information and communication technology (ICT) and measurement devices in rehabilitation engineering have been proposed. However, related studies have focused on analyzing motor responses to specific sensory stimuli, and the assessment measures often fail to capture a patient’s symptoms in dynamic environments. Therefore, this study proposes a methodology for modeling human spatial cognition. This cognitive architecture utilizes a structural coupling system that integrates parameters from multiple computational intelligence subsystems. In this study, we constructed a simulation environment capable of replicating the movements of patients with USN using empirical data collected from actual experiments. Furthermore, in this simulation environment, we developed patient agents that incorporated the proposed cognitive architecture. The experimental results suggest that hypothesis testing concerning attention mechanisms can be applied through the performance of patient agents within the simulation environment.

This study proposes two fuzzy clustering algorithms based on autoregressive moving average (ARMA) model for series data. The first, referred to as Tsallis entropy-regularized fuzzy c-ARMA model (TFCARMA), is created from k-ARMA, a conventional hard clustering algorithm for series data. TFCARMA is motivated by the relationship between the two clustering algorithms for vectorial data: k-means and Tsallis entropy-regularized fuzzy c-means. The second, referred to as q-divergence-based fuzzy c-ARMA model (QFCARMA), is created from ARMA mixtures, a conventional probabilistic clustering algorithm for series data. QFCARMA is motivated by the relationship between the two clustering algorithms for vectorial data: Gaussian mixture model and q-divergence-based fuzzy c-means. Based on numerical experiments using an artificial dataset, we observed the effects of fuzzification parameters in the proposed algorithms and relationship between the proposed and conventional algorithms. Moreover, numerical experiments using seven real datasets compared the clustering accuracy among the proposed and conventional algorithms.

Serendipity-oriented recommender systems aim to counteract the overspecialization of user preferences. However, evaluating a user’s serendipitous response to a recommended item can be challenging owing to its emotional nature. In this study, we address this issue by leveraging the rich knowledge of large language models (LLMs) that can perform various tasks. First, it explores the alignment between the serendipitous evaluations made by LLMs and those made by humans. In this study, a binary classification task was assigned to the LLMs to predict whether a user would find the recommended item serendipitously. The predictive performances of three LLMs were measured on a benchmark dataset in which humans assigned the ground truth to serendipitous items. The experimental findings revealed that LLM-based assessment methods do not have a very high agreement rate with human assessments. However, they performed as well as or better than the baseline methods. Further validation results indicate that the number of user rating histories provided to LLM prompts should be carefully chosen to avoid both insufficient and excessive inputs and that interpreting the output of LLMs showing high classification performance is difficult.

With the development of distant-water fisheries, ship fishing and fish catch detection are now vital to modern fishing. Existing manual detection methods are prone to issues such as missed detections and false detections. Deep learning has enabled the deployment of detection models on shipboard devices, offering a new solution. However, many existing models are hindered by large parameters and computational complexity, making them unsuitable for shipboard use due to limited resources and costs onboard ships. To address these challenges, we propose the RCT-YOLOv8 model for tuna catch detection in this paper. Specifically, we adopt YOLOv8 as the base model and replace the network backbone with RepVGG network, which employs re-parameterized convolutions to enhance detection accuracy. Additionally, we incorporate coordinate attention at the end of the backbone to better aggregate channel-wise information. In the neck part, we introduce the contextual transformer (CoT) attention and propose the C2F-CoT model, which combines convolutional neural network with Transformer to capture global features, thereby improving detection accuracy and the effectiveness of feature propagation. We test multiple loss functions and select efficient intersection over union, which is more suitable for our algorithm. Furthermore, to adapt to devices with limited computational resources, we utilize the dependency-graph-based pruning method to compress the network model. Compared to the base network, the pruned model achieves a 9.8% increase in detection accuracy while reducing parameters and computational complexity by 40% and 35.8%, respectively. Compared to various algorithms, the pruned model demonstrates the highest detection accuracy, lowest parameter count, and lowest computational complexity, achieving optimal results at all fronts.

The resolution and noise levels of input images directly affect the three-dimensional (3D) structure-from-motion (SfM) reconstruction performance. Conventional super-resolution (SR) methods focus too little on denoising, and latent image noise becomes worse when resolution is improved. This study proposes two SR denoising training algorithms to simultaneously improve resolution and noise: add-noise-before-downsampling and downsample-before-adding-noise. These portable methods preprocess low-resolution training images using real-world noise samples instead of altering the basic neural network. Hence, they concurrently improve resolution while reducing noise for an overall cleaner SfM performance. We applied these methods to the existing SR network: super-resolution convolutional neural network, enhanced deep residual super-resolution, residual channel attention network, and efficient super-resolution transformer, comparing their performances with those of conventional methods. Impressive peak signal-to-noise and structural similarity improvements of 0.12 dB and 0.56 were achieved on the noisy images of Smartphone Image Denoising Dataset, respectively, without altering the network structure. The proposed methods caused a very small loss (<0.01 dB) on clean images. Moreover, using the proposed SR algorithm makes the 3D SfM reconstruction more complete. Upon applying the methods to non-preprocessed and conventionally preprocessed models, the mean projection error was reduced by a maximum of 27% and 4%, respectively, and the number of 3D densified points was improved by 310% and 7%, respectively.

Named entity recognition (NER) usually focuses on general domains. Specific domains beyond the English language have rarely been explored. In Indonesian NER, the available resources for specific domains are scarce and on small scales. Building a large dataset is time-consuming and costly, whereas a small dataset is practical. Motivated by this circumstance, we contribute to specific-domain NER in the Indonesian language by providing a small-scale specific-domain NER dataset, IDCrossNER, which is semi-automatically created via automatic translation and projection from English with manual correction for realistic Indonesian localization. With the help of the dataset, we could perform the following analyses: (1) cross-domain transfer learning from general domains and specific-domain augmentation utilizing GPT models to improve the performance of small-scale datasets, and (2) an evaluation of supervised approaches (i.e., in- and cross-domain learning) vs. GPT-4o on IDCrossNER. Our findings include the following. (1) Cross-domain transfer learning is effective. However, on the general domain side, the performance is more sensitive to the size of the pretrained language model (PLM) than to the size and quality of the source dataset in the general domain; on the specific-domain side, the improvement from GPT-based data augmentation becomes significant when only limited source data and a small PLM are available. (2) The evaluation of GPT-4o on our IDCrossNER demonstrates that it is a powerful tool for specific-domain Indonesian NER in a few-shot setting, although it underperforms in prediction in a zero-shot setting. Our dataset is publicly available at https://github.com/khairunnisaor/idcrossner.

In semisupervised learning, particularly in dealing with health big data classification problems, optimizing the performance of classifiers has always been a challenge. Accordingly, this study explores an optimization algorithm based on collaborative training to better handle health big data. First, the tri-training and decision tree classification models were selected for comparison. The average classification accuracy of the tri-training classification model was 4.20% higher than that of the decision tree classification model. Subsequently, the standard tri-training classifier was compared with these two classifiers. The classification accuracy of the standard tri-training classifier increased by 3.88% and 4.33%, respectively, compared with the aforementioned two classifiers. Finally, under the condition of 10% labeled samples, the performance of the collaborative training optimization algorithm was verified under three different basis classifiers. The results of this study demonstrate the effectiveness of optimization algorithms based on collaborative training in dealing with health big data classification problems. By improving the performance of the classifier, health big data can be predicted and analyzed more accurately, thereby improving the accuracy and efficiency of medical decision-making. Meanwhile, the application of this optimization algorithm also provides new research directions for other semisupervised learning problems.

With the increasing focus on sustainable development in society, intelligent domain perception and digital twin technology can be used to evaluate and optimize the design of office furniture. This study analyzed sensor data through machine-learning and data-mining techniques to identify patterns and trends in the office environment. Simultaneously, a digital twin model of office furniture partition space was established to simulate the usage of furniture partition space throughout its full lifecycle. When 50% of nodes fail, the minimum transmission energy mode was significantly better than the maximum greedy forwarding mode in terms of cumulative throughput. The distributed, event-based, unsupervised clustering algorithm successfully reduced communication energy consumption, and the lightweight gradient boosting machine algorithm achieved the best design optimization rate, with an improvement of 0.53%. The ratio of value-added time to non-value-added time increased by 56.3%. The study aimed to provide innovative ideas for the development of intelligent office environments, promote the design of office furniture toward intelligence and sustainability, and improve the adaptability and efficiency of the work environment.

This study focuses on the stability and stabilization problems of a switching polynomial fuzzy system with a time-varying delay. The switching method is based on the operating sub-domains, and the switching polynomial Lyapunov function with a time-varying delay is used to design the switching polynomial static output feedback controller. The switching polynomial Lyapunov function contains a double-integral term for analyzing the upper bound of the time-varying delay. The stabilization of the polynomial system is investigated, using the boundary information of the membership functions and introducing slack polynomial matrices, which can reduce the conservatism of the stability conditions. Subsequently, the sum of squares conditions are obtained, which are convex and can be solved using SOSTOOLS. Finally, the viability and validity of the proposed approach are demonstrated using a numerical example.

This paper introduces a method for estimating 3D human joint angles using a hybrid optimization approach that integrates particle swarm optimization (PSO) with the steepest descent method for enhanced accuracy in both global and local searches. While advancements in motion capture technologies have made it easier to obtain 2D human joint position data, the accurate estimation of 3D joint angles remains crucial for detailed behavior analysis. Our proposed method first applies PSO to optimize the initial estimation of 3D joint angles from 2D joint positions. We further refine the estimation using the steepest descent method, improving the local search process. The convergence and accuracy of the algorithm are influenced by the grouping strategy in PSO, which is discussed in detail. Experimental results validate the effectiveness of our approach in enhancing the accuracy of 3D human pose estimation.

This paper presents an online topological mapping method on a quadcopter with fast-growing neural gas. Recently, perceiving the real world in 3D space has become increasingly important, and robotics is no exception. Quadcopters are the most common type of robot working in 3D space. The ability to perceive 3D space is even required in order to enable real-time autonomous control. Dense maps are simply unpractical, while sparse maps are not suitable due to a lack of appropriate information. Topological maps then offer a balance between computational cost and accuracy. One of the most well-known unsupervised learning methods for topological mapping is growing neural gas (GNG). Unfortunately, it is difficult to increase the learning speed due to the traditional iterative method. Consequently, we propose a new method for topological mapping, called simplified multi-scale batch-learning GNG, by applying a mini-batch strategy in the learning process. The proposed method has been implemented on a quadcopter for indoor mapping applications. In addition, the topological maps are also combined with the tracking data of the quadcopter to generate a new global map. The combination is simple yet robust, based on rotation and translation strategies. Thus, the quadcopter is able to run the algorithms in real-time and maintain its performance above 30 fps.

Skeleton-based human action recognition has great potential for human behavior analysis owing to its simplicity and robustness in varying environments. This paper presents a spatial and temporal attention-enhanced graph convolution network (STAEGCN) for human action recognition. The spatial-temporal attention module in the network uses convolution embedding for positional information and adopts multi-head self-attention mechanism to extract spatial and temporal attention separately from the input series of the skeleton. The spatial and temporal attention are then concatenated into an entire attention map according to a specific ratio. The proposed spatial and temporal attention module was integrated with an adaptive graph convolution network to form the backbone of STAEGCN. Based on STAEGCN, a two-stream skeleton-based human action recognition model was trained and evaluated. The model performed better on both NTU RGB+D and Kinetics 400 than 2s-AGCN and its variants. It was proven that the strategy of decoupling spatial and temporal attention and combining them in a flexible way helps improve the performance of graph convolution networks in skeleton-based human action recognition.

English grammar error correction (GEC) has been a popular topic over the past decade. The appropriateness of automatic evaluations, e.g., the combination of metrics and reference types, has been thoroughly studied for English GEC. Yet, such systematic investigations on the Chinese GEC are still insufficient. Specifically, we noticed that two representative Chinese GEC evaluation datasets, namely YACLC and MuCGEC, adopt fluency edits-based references with the automatic evaluation metric, which was designed for minimal edits-based references and differs from the convention of English GEC. However, it is unclear whether such evaluation settings are appropriate. Furthermore, we explored other dimensions of Chinese GEC evaluation, such as the number of references and tokenization granularity, and found that the two datasets exhibit significant differences. We hypothesize that these differences are crucial for Chinese GEC automatic evaluation. Thus, we publish the first human-annotated rankings on Chinese GEC system outputs and conducted an analytical meta-evaluation which discovered that 1) automatic evaluation metrics should match the types of reference; 2) the evaluation performance grows with the number of references, a consistent finding with English GEC, while four is the smallest reference number that empirically shows maximum correlation with human annotators; and 3) the granularity of tokenization has a minor impact, which is however a necessary preprocessing step for Chinese texts. We have made the proposed dataset publicly accessible at https://github.com/wang136906578/RevisitCGEC.