The Journal of Supercomputing最新文献

Porphyrazine is a macrocyclic molecule that has potential uses in biology and materials research. In this work, we investigate the topological characteristics of porphyrazine via topological indices. These indices are important in QSPR and QSAR modeling because they aid in the analysis and prediction of physical, biological, and chemical properties associated with a specific chemical structure. In this paper, we investigate neighborhood M-polynomial and graph index-entropy of porphyrazine graph, deriving several topological indices based on neighborhood degree sum from it, and numerical computation and graphical interpretation are used to explain the results further. This research advances our understanding of the basic principles of physics and chemistry by shedding light on the intricate connections between biological processes, chemical reactivity, and molecular structure.

Session-aware recommender systems analyze the sequential patterns of user actions to uncover the shifting preferences across sessions. User reviews enriched with sentiments can act as a guiding tool for session-aware systems. Existing methods for session-aware recommendations based on deep learning models do not consider the user’s sentiment granularity for generating reliable recommendations. In this paper, we have employed fuzzy-sentiment to guide the recommendation process toward a personalized and varied range of recommendations, resulting in an improved satisfaction level for the user. Fuzzy-sentiment provides a spectrum of sentiment scores (Highly positive, Positive, Neutral, Negative, and Highly Negative). This precise sentiment information allows the system to grasp the emotional tone and specific aspects of user experiences, shedding light on why users appreciated or were dissatisfied with a product. The sentiment scores are utilized to guide the recommendation process in the three-phase Sentiment-Guided Session-aware Recommender System, Fuzzy-SGSaRS. The first phase determines users’ sentiments from reviews about purchased products using the Fuzzy LSTM (FLSTM) technique. The learning process in the second phase employs a Graph Convolutional Network (GCN) to derive embeddings for Users, Interaction Sessions, and Products. The acquired embedding vectors are subsequently fed into the Double Deep Q-Network (DDQN) during the third phase to recommend intriguing products to the user(s). A series of experimental evaluations on four datasets of Amazon reviews illustrate that the proposed system outperformed various state-of-the-art methods.

Data centres have become the backbone of infrastructure for delivering cloud services. In the emerging cloud computing paradigm, virtual machine (VM) live migration involves moving a running VM across hosts without visible interruption to the client. Security vulnerabilities, resource optimization, and maintaining the quality of service are key issues in live VM migration. Maintaining security in VM live migration is one of the critical concerns. To create a secure environment, this paper proposes a live migration technique using the blowfish cryptographic algorithm for encryption and decryption, along with blockchain technology, to address challenges such as decentralization, data privacy, and VM security. The algorithms, namely key management blowfish encryption (KMBE), access control searchable encryption (ACSE), protected searchable destination server (PSDS), and key expansion blowfish decryption (KEBD), improve security in VM live migration in terms of various parameters such as data centre request servicing time, response time, and data transfer cost. The proposed technique KMBE improves migration cost ($) by 60–70%, ACSE reduces overall energy consumption (w) by 70–80%, PSDS reduces makespan (ms) by 40–50%, and KEBD improves the security in live VM migration by 30–40%.

It is a challenging task to achieve the minimum average CCT (coflow completion time) and provide isolation guarantees in multi-tenant datacenters without prior knowledge of coflow sizes. State-of-the-art solutions either focus on minimizing the average CCT or providing optimal isolation guarantees. However, achieving the minimum average CCT and isolation guarantees in multi-tenant datacenters is difficult due to the conflicting nature of these objectives. Therefore, we propose FIGCS-TF (Fast and Isolation Guarantees Coflow Scheduling via Traffic Forecasting), a coflow scheduling algorithm that does not require prior knowledge. FIGCS-TF utilizes a lightweight forecasting module to predict the relative scheduling priority of coflows. Moreover, it employs the MDRF (monopolistic dominant resource fairness) strategy for bandwidth allocation, which is based on super-coflows and helps achieve long-term isolation. Through trace-driven simulations, FIGCS-TF demonstrate communication stages that are 1.12(times), 1.99(times), and 5.50(times) faster than DRF (Dominant Resource Fairness), NCDRF (Non-Clairvoyant Dominant Resource Fairness) and Per-Flow Fairness, respectively. In comparison with the theoretically minimum CCT, FIGCS-TF experiences only a 46% increase in average CCT at the top 95th percentile of the dataset. Overall, FIGCS-TF exhibits superior performance in reducing average CCT compared to other algorithms.

Single-sample face and ear recognition (SSFER) is a challenging sub-problem in biometric recognition that refers to the difficulty in feature extraction and classification when only a single-face or ear training image is available. SSFER becomes much more challenging when images contain a variety of lighting, positions, occlusions, expressions, etc. Virtual sample generation methods in SSFER have gained popularity among researchers due to their simplicity in the augmentation of training sets and improved feature extraction. In this article, we propose a novel and simple method for the generation of virtual samples for training the classifiers to be used in SSFER. The proposed method is based on 2D local patches, and six training samples are generated for a single face or ear image. Further, training is performed using one of the variations along with its generated virtual samples, while during testing, all the variations were considered except the one used during training. Features are extracted using principal component analysis, and classification is performed using the nearest-neighbour classifier. Extensive experiments were performed for the image quality of the virtual samples, classification accuracy, and testing time on ORL, Yale, and AR (illumination) face databases, and AMI and IITD ear databases which are publicly available. The results are also compared with other state-of-the-art methods, with classification accuracy and universal image quality being the major outcomes. The proposed method improves the classification accuracy by 14.50%, 1.11%, 0.09%, 21.60%, and 10.00% on AR (illumination), Yale, ORL, IITD, and AMI databases, respectively. The proposed method showed an improvement in universal image quality by 15%, 20%, 14%, 30%, and 15% on AR (illumination), Yale, ORL, IITD, and AMI databases, respectively. Experimental results prove the effectiveness of the proposed method in generating virtual samples for SSFER.

Facial microexpressions (MEs) are involuntary, fleeting, and subtle facial muscle movements that reveal a person’s true emotional state and inner experiences. Microexpression recognition has been applied in various disciplines and fields, particularly in educational settings, where it can help educators better understand students’ emotional states and learning experiences, thus providing personalized teaching support and guidance. However, existing microexpression recognition datasets tailored for educational scenarios are limited. Moreover, microexpression recognition classifiers for educational settings not only require high recognition accuracy but also real-time performance. To this end, we provide a student behavior dataset specifically for research on microexpression and action recognition in educational scenarios. Moreover, we innovatively propose a lightweight dual-branch continuous attention network for microexpression recognition research. Specifically, for the student behavior dataset, we collect data on students” behaviors in real classroom scenarios. We categorize student microexpressions into two types: serious and non-serious. Additionally, we classify student classroom behaviors into several categories: attentive listening, note-taking, yawning, looking around, and nodding. Regarding the dual-branch continuous attention network, unlike most methods that extract features directly from entire video frames, which include abundant identity information, we focus on modeling subtle information from facial regions by using optical flow and motion information from keyframes as input. We extensively evaluate our proposed method on publicly available datasets such as CASME II and SAMM, as well as our provided dataset. The experimental results demonstrate that our proposed method achieves state-of-the-art performance in the field of microexpression recognition and provides a competitive dataset for analyzing student classroom behaviors in educational scenarios. We will provide the GitHub link upon acceptance of the paper, and we will make the dataset available to any applicant under a licensed agreement.

The accurate estimation of respiratory rate (RR) is crucial for assessing the respiratory system’s health in humans, particularly during auscultation processes. Despite the numerous automated RR estimation approaches proposed in the literature, challenges persist in accurately estimating RR in noisy environments, typical of real-life situations. This becomes especially critical when periodic noise patterns interfere with the target signal. In this study, we present a parallel driver designed to address the challenges of RR estimation in real-world environments, combining multi-core architectures with parallel and high-performance techniques. The proposed system employs a nonnegative matrix factorization (NMF) approach to mitigate the impact of noise interference in the input signal. This NMF approach is guided by pre-trained bases of respiratory sounds and incorporates an orthogonal constraint to enhance accuracy. The proposed solution is tailored for real-time processing on low-power hardware. Experimental results across various scenarios demonstrate promising outcomes in terms of accuracy and computational efficiency.

In the process of multi-attribute group decision-making (MAGDM), the cubic q-rung orthopair fuzzy sets (Cuq-ROFSs) are utilized to express membership and non-membership degrees in the form of interval values to efficiently cope with decision makers’ (DMs’) complex assessment values. To more efficiently capture DM evaluation results in the MAGDM procedure, we offer a novel tool called 2-tuple linguistic cubic q-rung orthopair fuzzy set (2TLCuq-ROFS), which extends Cuq-ROFS by using 2-tuple linguistic (2TL) terms. 2TLCuq-ROFS effectively incorporates the advantages of 2TL and Cuq-ROFS, making them attractive and versatile for depicting attribute values in an uncertain and complex decision-making environment. To effectively aggregate the attribute values in the form of 2-tuple linguistic cubic q-rung orthopair fuzzy numbers (2TLCuq-ROFNs), some Maclaurin symmetric mean (MSM) operators and their weighted forms are presented in this paper. The weight information for attributes is unknown. Therefore, the criteria importance through inter-criteria correlation (CRITIC) method is employed to determine the objective weight information. The purpose of this study is to incorporate a conventional multi-attributive border approximation area comparison (MABAC) framework based on 2TLCuq-ROFNs because it addresses problematic and imprecise decision-making problems by calculating the distance among each alternative and the border approximation area by using 2TLCuq-ROFNs and MSM aggregation operators. First, some basic concepts associated with 2TLCuq-ROFNs and the CRITIC-MABAC procedure are briefly explained. Moreover, an evaluation framework based on the improved CRITIC-MABAC method is established. An explanatory case study related to the risk investment problem in Belt and Road is used to verify the validity and practicality of the designed evaluation framework. In conclusion, by utilizing the CRITIC-MABAC methodology based on proposed operators, we find that (varLambda _7) is the optimal alternative for risk investment. Furthermore, comparison analysis emphasizes the integrity and prominent features of the proposed methodology and provides various complementary perspectives for investors.

The internet of things (IoT) is an exponentially growing network of physical objects equipped with sensors, software and network connectivities to collect, process, transmit and receive data. Wireless sensor networks (WSNs) play an essential role in supporting the IoT. These networks, made up of nodes with the ability to monitor their environment, enable the collection and transmission of specific data in real time, offering enhanced applications and services within IoT networks. This symbiosis between WSN and IoT can be defined as WSN-based IoT. The complexity of WSN-based IoT lies in the effective management of these varied devices, each with its own distinct capabilities. Clustering is a popular technique for reducing the communication load, conserving energy, aggregating data and optimizing the performance of WSN-based IoT systems. Once the cluster heads (CHs) are chosen, conventional clustering algorithms typically use a single radius of clustering (RC) to group devices into multiple clusters. However, this approach may not be optimized for WSN-based IoT networks, as devices may have different features, for example, the residual energy ((R_{rm Enrg})) and the distance to the base station (DistBS). In a previous work, we proposed the DCOPA (a distributed clustering based on objects performances aggregation for hierarchical communications in IoT applications) protocol for clustering in WSN-based IoT networks. DCOPA applies the same clustering algorithm to the elected CHs, without considering their distinctions in terms of (R_{rm Enrg}) and DistBS. The proposed new approach, called unequal-DCOPA (UDCOPA), allows us to define for each CH its adaptive radius of clustering (ARC) which will be sensitive to the local criteria of (R_{rm Enrg}) and DistBS of the CH concerned. The ARC is modeled as a multicriteria system applied to each CH. Simulation results show that our new UDCOPA approach outperforms DCOPA and LEACH protocols for energy management, load balancing, scalability and network lifetime. UDCOPA increases lifetime by (62.61%) over LEACH and by (32.72%) over DCOPA.

We propose a framework designed to tackle a multi-objective optimization challenge related to the placement of applications in fog computing, employing a deep reinforcement learning (DRL) approach. Unlike other optimization techniques, such as integer linear programming or genetic algorithms, DRL models are applied in real time to solve similar problem situations after training. Our model comprises a learning process featuring a graph neural network and two actor-critics, providing a holistic perspective on the priorities concerning interconnected services that constitute an application. The learning model incorporates the relationships between services as a crucial factor in placement decisions: Services with higher dependencies take precedence in location selection. Our experimental investigation involves illustrative cases where we compare our results with baseline strategies and genetic algorithms. We observed a comparable Pareto set with negligible execution times, measured in the order of milliseconds, in contrast to the hours required by alternative approaches.