The Journal of Supercomputing最新文献

In this work, we study the coupled system of fractional integro-differential equations, which includes the fractional derivatives of the Riemann–Liouville type and the fractional q-integral of the Riemann–Liouville type. We focus on the utilization of two significant fixed-point theorems, namely the Schauder fixed theorem and the Banach contraction principle. These mathematical tools play a crucial role in investigating the existence and uniqueness of a solution for a coupled system of fractional q-integro-differential equations. Our analysis specifically incorporates the fractional derivative and integral of the Riemann–Liouville type. To illustrate the implications of our findings, we present two examples that demonstrate the practical applications of our results. These examples serve as tangible scenarios where the aforementioned theorems can effectively address real-world problems and elucidate the underlying mathematical principles. By leveraging the power of the Schauder fixed theorem and the Banach contraction principle, our work contributes to a deeper understanding of the solutions to coupled systems of fractional q-integro-differential equations. Furthermore, it highlights the potential practical significance of these mathematical tools in various fields where such equations arise, offering a valuable framework for addressing complex problems.

In current researches on belief rule base (BRB), input parameters are tended to be expressed in the form of quantitative values through expert knowledge combined with optimization methods. A singular quantitative value fails to capture the statistical properties, leading to irrational outcomes. Therefore, an attempt on attribute weights is made in this paper, and a new model with probability distribution attribute weights (pdw) called BRB-pdw is proposed. The combination of two attributes is in detail discussed, where attribute weights are described as random variables with specific probability distribution. To characterize the output of probability distribution attribute weight, a new concept of expectation of activation weight is proposed. In addition, the BRB-pdw is extended to multiple attributes to demonstrate its universality. Furthermore, fundamental properties and characteristics of the BRB-pdw are further validated by rigorous mathematical derivation. Finally, practicability of the BRB-pdw is validated with NASA lithium battery open dataset, and experiments show that the BRB-pdw model is more robust while maintaining precision.

The existing I²C interfaces require significant CPU intervention for data communication. In SOC systems, when using internally integrated I²C modules as masters, software control of I/O ports is necessary to emulate the I²C protocol for data transmission. This paper proposes a Mini I²C bus interface circuit design scheme that supports both master and slave modes. The I²C interface features minimal CPU intervention during data transmission, ease of use, small circuit area, and low power consumption. Additionally, the internal state machine design employs independent finite state machines (FSMs) for master and slave modes, enabling flexible configuration of the I²C module to operate in either mode. In comparison to reference (ShenZhen in Microelectronics Technology CO.BJ8M306A, Datasheet.2019.12.2., 2019), the proposed I²C solution reduces CPU instructions by 50% during data transmission, and by 33% compared to reference (GigaDevice Semiconductor Inc. GD32F1x0, Datasheet, 2022). After DC synthesis, the proposed design occupies only 14% of the area and consumes only 3.6% of the power of the Open Source I²C Design (Forencich in verilog-i2c. GitHub repository. Retrieved from https://github.com/alexforencich/verilog-i2c, n.d.). Therefore, this design scheme is better suited for low-power systems. The proposed design was validated through simulation using Xilinx ISE 14.7 with a SPARTAN 3 FPGA model xc3s500e-5pq208, and finally implemented using Huahong 95 nm CMOS technology, demonstrating high integration and low power consumption.

In pattern recognition, statistics, machine learning, and data mining, feature or attribute selection is a standard dimensionality reduction method. The goal is to apply a set of rules to select essential and relevant features from the original dataset. In recent years, unsupervised feature selection approaches have garnered significant attention across various research fields. This study presents a well-organized summary of the latest and most effective unsupervised feature selection techniques in the scientific literature. We introduce a taxonomy of these strategies, elucidating their significant features and underlying principles. Additionally, we outline the pros, cons, challenges, and practical applications of the broad categories of unsupervised feature selection approaches reviewed in the literature. Furthermore, we conducted a comparison of several state-of-the-art unsupervised feature selection methods through experimental analysis.

Multimodal named entity recognition (MNER) is an emerging foundational task in natural language processing. However, existing methods have two main limitations: 1) previous methods have focused on the visual representation of the entire image or target objects. However, they overlook the fine-grained semantic correspondence between entities and visual target objects, or ignore the visual cues of the overall scene and background details in the image. 2) Existing methods have not effectively overcome the semantic gap between different modalities due to the heterogeneity between text and images. To address these issues, we propose a novel multimodal heterogeneous graph entity-level fusion method for MNER (HGMVG) to achieve cross-modal feature interaction from coarse to fine between text and images under the guidance of visual information at different granularities, which can improve the accuracy of named entity recognition. Specifically, to resolve the first issue, we cascade cross-modal semantic interaction information between text and vision at different visual granularities to obtain a comprehensive and effective multimodal representation. For the second issue, we describe the precise semantic correspondences between entity-level words and visual target objects via multimodal heterogeneous graphs, and utilize heterogeneous graph attention networks to achieve cross-modal fine-grained semantic interactions. We conduct extensive experiments on two publicly available Twitter datasets, and the experimental results demonstrate that HGMVG outperforms the current state-of-the-art models in the MNER task.

The proper positioning of Virtual Machines (VMs) on the hosts in a cloud environment reduces the need for the VM migration and its consequences. The positioning becomes more significant when there exists a multi-cloud environment where the hosts exist on multi-site datacenters. Based on user’s requests, VMs should be dynamically positioned; however, if the users’ future demands can be predicted, the positioning can be adaptively done in advance, which is both more cost-effective for users and more requests are met. To this end, at the request of their users, VMs’ providers can reserve VMs for the users’ future needs. However, if some users would not like to reserve VMs or if the number of reserved VMs is less than users’ needs, VMs should be allocated on demand. However, the reserve or on-demand policy cannot be applied freely if users have constraints and objectives. Among others, cost of using resources and response time are the most important users’ objectives, and load balancing hosts and datacenters for the proper resource utilization is the most important providers’ objective. To consider the reserve policy, a multi-layered model is presented in this paper where a multi-objective optimization is used to meet the objectives. The proposed model was applied to Google, Wikipedia, and NASA datasets. The results show: (1) The number of predicted VMs for reserve is closer to the real VMs requested in datasets NASA, Wikipedia, and Google than the related work. This was due to the use of a dynamic neural network, called NARX; (2) objective cost is regarded more than the related work, while it respects more trade-off between the user’s objectives and provider’s one; (3) placement of VMs on hosts is done in a balanced way, leading to the reduction of overloaded hosts and response time.

Clinical images of brain tumors (BT) are crucial in the diagnostic process and contain substantial medical information. In neurosurgery and neurology, AI’s application in retrieving and analyzing brain tumors leads to earlier, more accurate diagnoses and improves treatment planning. However, the accuracy of the existing methods for the physical retrieval of similar images needs to be improved. This paper introduces Gradient Ladybug Beetle Optimization-based LeNet (GLBO-LeNet) for the retrieval of brain tumor magnetic resonance images (MRI) from the medical datasets. This approach processes both input MRI images and query MRIs using the same pipeline. Tumor segmentation process is performed on these images using a 3D Convolutional Neural Network (CNN). Features are extracted from segmented images, incorporating a novel feature extraction method, LTDP based on Discrete Wavelet Transform (DWT) with Pyramid Histogram of Orientation (PHoG). The extracted features are utilized for tumor classification using LeNet-5, tuned by Gradient Ladybug Beetle Optimization (GLBO). The classified outputs from input MRI images are indexed in an image database. Similar images are retrieved and ranked using a proposed hybrid similarity measure, enabling efficient brain MRI image retrieval. In this study, the GLBO-LeNet-based brain tumor MRI retrieval system achieved an accuracy of 91.5%, a Prue-positive rate (TPR) of 91.9%, a True-negative rate (TNR) of 92.5%, a Positive predictive value (PPV) of 90.8% and a Negative predictive value (NPV) of 89.4%.

Ongoing research is focused on underwater wireless sensor networks (UWSNs), which hold great promise for various applications to improve human lives. Many sensor nodes have been strategically positioned in rivers and oceans to closely monitor and analyze the underwater ecosystem. UWSNs face significant challenges due to energy limitations, unreliable communication channels, hole occurrence, and low packet delivery ratio. This paper introduces effective energy routing, hole detection and mitigation, and reliable delivery schemes to address these issues. In ERR-UWSN, a single-path routing strategy optimizes energy consumption by selecting the best-sending node through a unique weight function. However, the reliability of this approach is compromised in the harsh and unpredictable submarine environment. To overcome this limitation, co-ERR-UWSN combines cooperative routing with relay nodes between source–destination pairs. A carefully selected weight function ensures low energy consumption, reliability, and optimal node selection for data transmission. The simulation results show that ERR-UWSN is better than CEER, Co-EPRR, and FB-DBR in terms of residual energy, end-to-end delay, energy consumption, and total number of alive nodes. Also, Co-ERR-UWSN has a special advantage in the PDR criterion compared to other schemes.

The super edge-connectivity of a connected graph G, denoted by ({lambda }'left( G right) ), if exists, is the minimum number of edges whose deletion disconnects the graph such that each component has no isolated vertices. The direct product of graphs G and H, denoted by (Gtimes H), is the graph with vertex set (Vleft( Gtimes H right) =Vleft( G right) times Vleft( H right) ), where two vertices (left( {{u}_{1}},{{v}_{1}} right) ) and (left( {{u}_{2}},{{v}_{2}} right) ) are adjacent in (Gtimes H) if and only if ({{u}_{1}}{{u}_{2}}in Eleft( G right) ) and ({{v}_{1}}{{v}_{2}}in Eleft( H right) ). In this paper, it is proved that ({lambda }'left( Gtimes {{C}_{n}} right) = min { 2n{lambda }'left( G right) ,2underset{xyin Eleft( G right) }{{min }},left( {{deg }_{G}}left( x right) +{{deg }_{G}}left( y right) right) -2 }) for (i) any connected graph G with (left| G right| le n) or (Delta left( G right) le n-1) and an odd cycle ({{C}_{n}}), or (ii) any split graph G with (left| G right| le n) or (Delta left( G right) le n-1) and a cycle ({{C}_{n}}).

Software evolution measurement is required to control software costs and aid in the development of cost-effective software. Early detection of potential changes gives developers time to plan for change. Simple techniques to detect the change-proneness of classes are required such as thresholds, particularly in incremental software development. In this study, we propose to derive thresholds to detect the change-proneness of classes using ROC analysis. The analysis is conducted on the evolution of five systems for six object-oriented metrics, Chidamber and Kemerer. Thresholds are considered in software evolution in three intervals: 6 months, 12 months, and 3 years. Thresholds are reported for four metrics that can predict change-proneness. Similar thresholds are reported at 6 and 12 months. For the same metrics, fault-proneness thresholds are identified, and the results are compared to their counterparts in change-proneness thresholds. The change-proneness thresholds derived are smaller and identify more classes for further investigation.