Vol. 43 · No. 6

In recent years, the utilization of Cloud computing services has significantly increased, placing a heightened demand on the workload, computational infrastructure, storage, and communication networks managed by datacenters. This surge has prompted researchers to enhance Cloud performance, focusing on minimizing execution time and computational costs. To address this challenge, efficient scheduling strategies involve the allocation of Cloud workload across different datacenters through the migration of virtual machines. This article examines and compares three methodologies to tackle this issue: one based on an Adaptive Neuro-Fuzzy Inference System, another utilizing a Swarm Fuzzy System, and a third one employing a Genetic Algorithm. The study evaluates their effectiveness for workflow scheduling using a CloudSim-based simulator in terms of makespan and computational costs. Results reveal that the neuro-fuzzy system outperforms the fuzzy and genetic systems regarding makespan in Montage and CyberShake environments. It demonstrates a computational cost advantage, achieving reductions of 7.01

The existing modeling methods for business process management systems (BPMS) focus on the logical and abstract data layers but often ignore operations related to the underlying database, thus failing to describe system behavior accurately. Workflow nets with tables (WFT-nets) compensate for the lack of description of database table operations in existing modeling methods. However, for timed business process management systems (TBPMS), their correct behavior depends on the logical correctness of the results obtained by the operational process and the time required for each activity to be correctly executed within a specified time. Since WFT-nets do not consider time properties, they are unable to describe time-related activities in TBPMS, potentially leading to incorrect results. This paper introduces timed workflow nets with tables (TWFT-nets), which add time elements to transitions to simulate time-constrained activities in the system. Additionally, we assign different labels to represent the execution strategies of activities under different time constraints. To analyze the soundness of TWFT-nets, we propose a timed database computation tree logic (TDCTL) model checking method and define soundness from three perspectives: logic control layer, data design requirements, and time constraints. We transform soundness into TDCTL formulas, provide a model checking algorithm, and develop a tool. Experiments show the effectiveness of our methods.

Hot-rolled steel strip plays an important role in the field of industrial manufacturing. In addition, defects on its surface affect the aesthetics of the subsequent products and their corrosion resistance, wear resistance, and fatigue strength. However, the existing methods are difficult to learn or capture discriminative feature representations, resulting in poor detection performance. Therefore, its surface defect detection faces two main challenges: one is the insufficient ability to extract local features, and the other is the limited ability to detect multi-scale targets. To address the above issues, we propose a Residual Deformable Convolution and Double LayerNorm Swin Transformer and Channel Expansion Feature Pyramid Networks (RTCN) multi-scale hot-rolled strip surface defect detection model, which adopts Double LayerNorm Swin Transformer (DLST) and as Residual Deformable Convolution Block (RDCB) its backbone network to increase the sensitivity of the model's detection of small and irregular defects. In addition, we adopt Channel Expansion Feature Pyramid Networks (CEFPN) to introduce more feature dimensions to better capture the structure and semantic image information. Ultimately, we assess the proposed model using the publicly available NEU-DET dataset. Our comprehensive testing shows that the model developed in this paper beats the most advanced approach by 1.1

Brain-computer interface (BCI) technology holds immense promise in the rehabilitation of patients with movement disorders, leveraging the body's physiological mechanisms to enhance their quality of life by reshaping motor neural circuits through external devices. Nevertheless, current BCI applications for rehabilitation predominantly rely on a single physiological signal, often overlooking the synergistic impact of multiple signals. Simultaneously, while reinforcement learning shows significant potential for BCI applications, there exists a scarcity of studies exploring this intersection.This paper introduces a novel motor intention judgment model grounded in multimodal signal fusion and reinforcement feature selection. The model adeptly extracts comprehensive motor intention features by integrating pertinent information from both electroencephalogram (EEG) and electromyogram (EMG). Furthermore, reinforcement learning is employed for judicious feature selection, yielding promising outcomes in subsequent experiments. The study utilizes publicly available datasets to diagnose the motion intention of the subjects, complemented by ablation experiments to affirm the efficacy of the model components. In instances of feature-level fusion, the model demonstrated a noteworthy enhancement in the average five-classification accuracy, surpassing results obtained from isolated EEG and EMG experiments by 28.46

Entity matching (EM) is a critical yet complex component of data cleaning and integration. Recent advancements in EM have predominantly been driven by deep learning (DL) methods. These methods primarily enhance data accuracy within structured data that adheres to a high-quality and well-defined schema. However, these schema-centric DL strategies struggle with the semantic web's linked data, which tends to be voluminous, semi-structured, diverse, and often noisy. To tackle this, we introduce a novel approach that is loosely schema-aware and leverages cutting-edge developments in DL, specifically transformer neural networks and pre-trained language models. We evaluated our approach on six datasets, including two tabular and four RDF datasets from the semantic web. The findings demonstrate the effectiveness of our model in managing the complexities of noisy and varied data.

We present a cloud management service called RAIN. It has been designed specifically for versatility and scalability of operation, allowing for the processing of a large number of requests at the same time. Its operation is transactional and controlled by a workflow of operations forming one requisition. Requisitions and their operations can be executed in parallel, allowing for high throughput and scalability of the controlled cloud environment(s). The service is being used in day-to-day operations in a commercial environment. It is also designed for high failure tolerance, which is necessary when operating on third party cloud infrastructures. It has been developed and actively used for several years now, giving us a mature tool with many important features added over time, allowing for practical day-to-day operations. The architecture of the service is open and easily extendable to allow the inclusion of new cloud services of various types – PaaS providers as well as providers of higher-level services. The service is accessed via an asynchronous REST API. It allows the caller to resume execution and not wait for cloud deployment operations to take an arbitrary amount of time to finish, receiving progress updates via a simple callback REST API.

To solve the problem of time-consuming and low efficiency in manual defect detection, this paper proposes a bonding defect detection algorithm based on improved Single Shot MultiBox Detector (SSD). DenseNet is used to replace VGG of the SSD algorithm to improve the detection effect of bonding defect. A novel feature fusion network is designed, in which dilated convolution is used to reduce the size of the low-level feature map, and it is fused with the high-level feature map, and then the Convolutional Block Attention Module (CBAM) attention mechanism is used to increase the ability to extract the features. Focal loss is used to control the ratio of positive and negative samples for training and suppress easily separable samples, so that the samples involved in training have better distribution and the model has better detection performance. Then, the defect data set is constructed and a comparison experiment is carried out. The results show that the mAP, Precision, and Recall of the improved SSD network are increased to 75.9

Ultrasound examination is of great significance in the clinical diagnosis of diseases. Processing and analyzing ultrasound images through artificial intelligence technology and providing assistance in decision-making has been a hot topic of research for several years. However, since most medical images exist in the form of pictures, the current processing methods for ultrasound images basically continue to adopt the technical achievements related to static medical image processing not considering the characteristics reflected by the dynamically changing ultrasound images thus resulting in a missed diagnosis of diseases. To this end, this paper proposes an innovative multi-feature extraction and fusion method for ultrasound dynamic image classification which first extracts various types of underlying features such as texture, edge, and shape of salient targets in medical images that apply to dynamic images. Then, the feature frequency-inverse image frequency (FF-IIF) multi-feature fusion algorithm is used to generate an adaptive combined feature classification. In the experiments, the effects of the proposed algorithm are verified for three ultrasound examination items respectively. The experimental results show that the features extracted by the multi-feature fusion algorithm using FF-IIF still maintain a certain degree of fault tolerance and stability under the dynamic changes of ultrasound probe position and orientation. The computation time of the algorithm is moderate and perfectly adapted to the real-time examination of ultrasound medicine.

Currently, renewable energy sources (RES) have been widely deployed in the smart grid, especially in the active distribution network (ADN). However, the inherent uncertainty of the renewable energy output significantly impacts the economy of the ADN operation. It is suggested that the utilization of flexible resources (FR) can effectively even out the uncertainty of RES. Nevertheless, the non-marketization of FR may prevent the emerging park-level integrated energy systems (PIES), important entities in ADN, from sufficiently offering their potential flexibilities. Therefore, this paper presents a two-stage local flexibility trading mechanism for motivating multi-PIESs to provide their flexible resources (PFR) to improve the operational flexibility of ADN. The first stage determines the dispatching plan of ADN according to the day-ahead forecasted values of RES. The second stage enables the multi-PIESs to trade their PFR with ADN to adjust the day-ahead dispatching plan in real-time to correct the forecasted errors of RES. In terms of implementing the two stages, firstly, the capacity of PFR with involving the adjustable tie line power of PIES is quantified using an optimization-based assessment model. Secondly, based on the change of the operation cost before and after the sale of PFR, a pricing model of PFR is established. And then, to determine the trading amount of PFR, a real-time ADN economic dispatching model with the network constraints is further constructed, which aims at minimizing the comprehensive operation cost. Afterwards, a marginal-based method is employed to obtain the clearing price of PFR. Finally, to ensure the feasibility of the trading results and to provide the accurate dispatching strategies for the PFR trading in next time interval, a rolling dispatch for the multi-PIESs is carried out. Case studies demonstrate that the presented flexibility trading mechanism significantly reduces the power curtailment of ADN, the operation costs of ADN and the multi-PIESs.

Tourist attraction recommendation algorithms have been developed to meet demand related to tourism, spiritual and cultural pursuits. While many studies have been conducted on such algorithms, problems remain regarding tourist interest point recommendation such as ignoring social information, underutilizing context information, and not capturing node relationships which have limited the recommendation performance and representation capability. This paper proposes an algorithm based on graph convolutional neural networks and collaborative filtering (GCNs-CF) for travel interest point recommendation, using an image denoising encoder (IDE) instead of domain aggregation, to better capture the relationships and features between users and adjacent nodes of travel interest point nodes. An adaptive adjustment of the negative sample gradient size is used to solve the problem of slow convergence of graph convolutional neural network. The experimental results show that the proposed method has a higher recommendation effect than other algorithms.