CAAI Transactions on Intelligence Technology最新文献

Modern automated generation control (AGC) is increasingly complex, requiring precise frequency control for stability and operational accuracy. Traditional PID controller optimisation methods often struggle to handle nonlinearities and meet robustness requirements across diverse operational scenarios. This paper introduces an enhanced strategy using a multi-objective optimisation framework and a modified non-dominated sorting genetic algorithm II (SNSGA). The proposed model optimises the PID controller by minimising key performance metrics: integration time squared error (ITSE), integration time absolute error (ITAE), and rate of change of deviation (J). This approach balances convergence rate, overshoot, and oscillation dynamics effectively. A fuzzy-based method is employed to select the most suitable solution from the Pareto set. The comparative analysis demonstrates that the SNSGA-based approach offers superior tuning capabilities over traditional NSGA-II and other advanced control methods. In a two-area thermal power system without reheat, the SNSGA significantly reduces settling times for frequency deviations: 2.94s for $��\Delta �f_1�$ and 4.98s for $��\Delta �f_2�$, marking improvements of 31.6% and 13.4% over NSGA-II, respectively.

Urban traffic prediction with high precision is always the unremitting pursuit of intelligent transportation systems and is instrumental in bringing smart cities into reality. The fundamental challenges for traffic prediction lie in the accurate modelling of spatial and temporal traffic dynamics. Existing approaches mainly focus on modelling the traffic data itself, but do not explore the traffic correlations implicit in origin-destination (OD) data. In this paper, we propose STOD-Net, a dynamic spatial-temporal OD feature-enhanced deep network, to simultaneously predict the in-traffic and out-traffic for each and every region of a city. We model the OD data as dynamic graphs and adopt graph neural networks in STOD-Net to learn a low-dimensional representation for each region. As per the region feature, we design a gating mechanism and operate it on the traffic feature learning to explicitly capture spatial correlations. To further capture the complicated spatial and temporal dependencies among different regions, we propose a novel joint feature, learning block in STOD-Net and transfer the hybrid OD features to each block to make the learning process spatiotemporal-aware. We evaluate the effectiveness of STOD-Net on two benchmark datasets, and experimental results demonstrate that it outperforms the state-of-the-art by approximately 5% in terms of prediction accuracy and considerably improves prediction stability up to 80% in terms of standard deviation.

Cross-domain graph anomaly detection (CD-GAD) is a promising task that leverages knowledge from a labelled source graph to guide anomaly detection on an unlabelled target graph. CD-GAD classifies anomalies as unique or common based on their presence in both the source and target graphs. However, existing models often fail to fully explore domain-unique knowledge of the target graph for detecting unique anomalies. Additionally, they tend to focus solely on node-level differences, overlooking structural-level differences that provide complementary information for common anomaly detection. To address these issues, we propose a novel method, Synthetic Graph Anomaly Detection via Graph Transfer and Graph Decouple (GTGD), which effectively detects common and unique anomalies in the target graph. Specifically, our approach ensures deeper learning of domain-unique knowledge by decoupling the reconstruction graphs of common and unique features. Moreover, we simultaneously consider node-level and structural-level differences by transferring node and edge information from the source graph to the target graph, enabling comprehensive domain-common knowledge representation. Anomalies are detected using both common and unique features, with their synthetic score serving as the final result. Extensive experiments demonstrate the effectiveness of our approach, improving an average performance by 12.6$��\%�$ on the AUC-PR compared to state-of-the-art methods.

Lexical analysis is a fundamental task in natural language processing, which involves several subtasks, such as word segmentation (WS), part-of-speech (POS) tagging, and named entity recognition (NER). Recent works have shown that taking advantage of relatedness between these subtasks can be beneficial. This paper proposes a unified neural framework to address these subtasks simultaneously. Apart from the sequence tagging paradigm, the proposed method tackles the multitask lexical analysis via two-stage sequence span classification. Firstly, the model detects the word and named entity boundaries by multi-label classification over character spans in a sentence. Then, the authors assign POS labels and entity labels for words and named entities by multi-class classification, respectively. Furthermore, a Gated Task Transformation (GTT) is proposed to encourage the model to share valuable features between tasks. The performance of the proposed model was evaluated on Chinese and Thai public datasets, demonstrating state-of-the-art results.

Extrapolation on Temporal Knowledge Graphs (TKGs) aims to predict future knowledge from a set of historical Knowledge Graphs in chronological order. The temporally adjacent facts in TKGs naturally form event sequences, called event evolution patterns, implying informative temporal dependencies between events. Recently, many extrapolation works on TKGs have been devoted to modelling these evolutional patterns, but the task is still far from resolved because most existing works simply rely on encoding these patterns into entity representations while overlooking the significant information implied by relations of evolutional patterns. However, the authors realise that the temporal dependencies inherent in the relations of these event evolution patterns may guide the follow-up event prediction to some extent. To this end, a Temporal Relational Context-based Temporal Dependencies Learning Network (TRenD) is proposed to explore the temporal context of relations for more comprehensive learning of event evolution patterns, especially those temporal dependencies caused by interactive patterns of relations. Trend incorporates a semantic context unit to capture semantic correlations between relations, and a structural context unit to learn the interaction pattern of relations. By learning the temporal contexts of relations semantically and structurally, the authors gain insights into the underlying event evolution patterns, enabling to extract comprehensive historical information for future prediction better. Experimental results on benchmark datasets demonstrate the superiority of the model.

The cloud data centres evolved with an issue of energy management due to the constant increase in size, complexity and enormous consumption of energy. Energy management is a challenging issue that is critical in cloud data centres and an important concern of research for many researchers. In this paper, we proposed a cuckoo search (CS)-based optimisation technique for the virtual machine (VM) selection and a novel placement algorithm considering the different constraints. The energy consumption model and the simulation model have been implemented for the efficient selection of VM. The proposed model CSOA-VM not only lessens the violations at the service level agreement (SLA) level but also minimises the VM migrations. The proposed model also saves energy and the performance analysis shows that energy consumption obtained is 1.35 kWh, SLA violation is 9.2 and VM migration is about 268. Thus, there is an improvement in energy consumption of about 1.8% and a 2.1% improvement (reduction) in violations of SLA in comparison to existing techniques.

Real-time water-medium endoscopic images can assist doctors in performing operations such as tissue cleaning and nucleus pulpous removal. During medical operating procedures, it is inevitable that tissue particles, debris and other contaminants will be suspended within the viewing area, resulting in blurred images and the loss of surface details in biological tissues. Currently, few studies have focused on enhancing such endoscopic images. This paper proposes a water-medium endoscopic image processing method based on dual transmittance in accordance with the imaging characteristics of spinal endoscopy. By establishing an underwater imaging model for spinal endoscopy, we estimate the transmittance of the endoscopic images based on the boundary constraints and local image contrast. The two transmittances are then fused and combined with transmittance maps and ambient light estimations to restore the images before attenuation, ultimately enhancing the details and texture of the images. Experiments comparing classical image enhancement algorithms demonstrate that the proposed algorithm could effectively improve the quality of spinal endoscopic images.

Data augmentation plays an important role in boosting the performance of 3D models, while very few studies handle the 3D point cloud data with this technique. Global augmentation and cut-paste are commonly used augmentation techniques for point clouds, where global augmentation is applied to the entire point cloud of the scene, and cut-paste samples objects from other frames into the current frame. Both types of data augmentation can improve performance, but the cut-paste technique cannot effectively deal with the occlusion relationship between the foreground object and the background scene and the rationality of object sampling, which may be counterproductive and may hurt the overall performance. In addition, LiDAR is susceptible to signal loss, external occlusion, extreme weather and other factors, which can easily cause object shape changes, while global augmentation and cut-paste cannot effectively enhance the robustness of the model. To this end, we propose Syn-Aug, a synchronous data augmentation framework for LiDAR-based 3D object detection. Specifically, we first propose a novel rendering-based object augmentation technique (Ren-Aug) to enrich training data while enhancing scene realism. Second, we propose a local augmentation technique (Local-Aug) to generate local noise by rotating and scaling objects in the scene while avoiding collisions, which can improve generalisation performance. Finally, we make full use of the structural information of 3D labels to make the model more robust by randomly changing the geometry of objects in the training frames. We verify the proposed framework with four different types of 3D object detectors. Experimental results show that our proposed Syn-Aug significantly improves the performance of various 3D object detectors in the KITTI and nuScenes datasets, proving the effectiveness and generality of Syn-Aug. On KITTI, four different types of baseline models using Syn-Aug improved mAP by 0.89%, 1.35%, 1.61% and 1.14% respectively. On nuScenes, four different types of baseline models using Syn-Aug improved mAP by 14.93%, 10.42%, 8.47% and 6.81% respectively. The code is available at https://github.com/liuhuaijjin/Syn-Aug.

Although the image dehazing problem has received considerable attention over recent years, the existing models often prioritise performance at the expense of complexity, making them unsuitable for real-world applications, which require algorithms to be deployed on resource constrained-devices. To address this challenge, we propose WaveLiteDehaze-Network (WLD-Net), an end-to-end dehazing model that delivers performance comparable to complex models while operating in real time and using significantly fewer parameters. This approach capitalises on the insight that haze predominantly affects low-frequency information. By exclusively processing the image in the frequency domain using discrete wavelet transform (DWT), we segregate the image into high and low frequencies and process them separately. This allows us to preserve high-frequency details and recover low-frequency components affected by haze, distinguishing our method from existing approaches that use spatial domain processing as the backbone, with DWT serving as an auxiliary component. DWT is applied at multiple levels for better information retention while also accelerating computation by downsampling feature maps. Subsequently, a learning-based fusion mechanism reintegrates the processed frequencies to reconstruct the dehazed image. Experiments show that WLD-Net outperforms other low-parameter models on real-world hazy images and rivals much larger models, achieving the highest PSNR and SSIM scores on the O-Haze dataset. Qualitatively, the proposed method demonstrates its effectiveness in handling a diverse range of haze types, delivering visually pleasing results and robust performance, while also generalising well across different scenarios. With only 0.385 million parameters (more than 100 times smaller than comparable dehazing methods), WLD-Net processes 1024 × 1024 images in just 0.045 s, highlighting its applicability across various real-world scenarios. The code is available at https://github.com/AliMurtaza29/WLD-Net.

Early exiting has shown significant potential in accelerating the inference of pre-trained language models (PLMs) by allowing easy samples to exit from shallow layers. However, existing early exiting methods primarily rely on local information from individual samples to estimate prediction uncertainty for making exiting decisions, overlooking the global information provided by the sample population. This impacts the estimation of prediction uncertainty, compromising the reliability of exiting decisions. To remedy this, inspired by principal component analysis (PCA), the authors define a residual score to capture the deviation of features from the principal space of the sample population, providing a global perspective for estimating prediction uncertainty. Building on this, a two-stage exiting strategy is proposed that integrates global information from residual scores with local information from energy scores at both the decision and feature levels. This strategy incorporates three-way decisions to enable more reliable exiting decisions for boundary region samples by delaying judgement. Extensive experiments on the GLUE benchmark validate that the method achieves an average speed-up ratio of 2.17× across all tasks with minimal performance degradation. Additionally, it surpasses the state-of-the-art E-LANG by $��11�\%�$ in model acceleration, along with a performance improvement of 0.6 points, demonstrating a better performance-efficiency trade-off.