Computing最新文献

Hybrid networks, benefiting from both CNNs and Transformers architectures, exhibit stronger feature extraction capabilities compared to standalone CNNs or Transformers. However, in hybrid networks, the lack of attention in CNNs or insufficient refinement in attention mechanisms hinder the highlighting of target regions. Additionally, the computational cost of self-attention in Transformers poses a challenge to further improving network performance. To address these issues, we propose a novel point-to-point Dynamic Attention Guider(DAG) that dynamically generates multi-scale large receptive field attention to guide CNN networks to focus on target regions. Building upon DAG, we introduce a new hybrid network called the Dynamic Attention Guider Network (DAGN), which effectively combines Dynamic Attention Guider Block (DAGB) modules with Transformers to alleviate the computational cost of self-attention in processing high-resolution input images. Extensive experiments demonstrate that the proposed network outperforms existing state-of-the-art models across various downstream tasks. Specifically, the network achieves a Top-1 classification accuracy of 88.3% on ImageNet1k. For object detection and instance segmentation on COCO, it respectively surpasses the best FocalNet-T model by 1.6 (AP^b) and 1.5 (AP^m), while achieving a top performance of 48.2% in semantic segmentation on ADE20K.

With the development of IoT, the concept of intelligent services has gradually come to the fore. Intelligent services usually involve a large number of computation intensive tasks with data dependencies that are often modelled as directed acyclic graphs (DAGs), and the offloading of DAG tasks is complex and has proven to be an NP hard challenge. As a key research issue, the task offloading process migrates the computation intensive tasks from resource-constrained IoT devices to nearby edge servers, and pursuing a lower delay and energy consumption. However, data dependencies among tasks are complex, and it is challenging to coordinate the computation intensive tasks among multiple edge servers. In this paper, a flexible and generic DAG task model is built to support the associative task offloading process with complex data dependencies in IoT edge computing environments. Additionally, a priority-based DAG task offloading algorithm and a secondary resource allocation algorithm are proposed to minimize the response delay and improve the resource utilization of edge servers. Experimental results demonstrate that the proposed method can well support the DAG task offloading process with the shortest response delay, while outperforming all the benchmark policies, which is suitable for IoT edge computing environments.

The growing popularity of blockchains highlights the need to improve their scalability. While previous research has focused on scaling transaction processing, the scalability of transaction creation remains unexplored. This issue is particularly important for organizations needing to send large volumes of transactions quickly or continuously. Scaling transaction creation is challenging, especially for blockchain platforms like Ethereum, which require transactions to include a sequence number. This paper proposes four different methods to scale transaction creation. Our experimental evaluation assesses the scalability and latency of these methods, identifying two as feasible for scaling transaction creation. Additionally, we provide an in-depth theoretical analysis of these two methods.

Lead generation refers to the identification of potential topics (the ‘leads’) of importance for journalists to report on. In this article we present (mu )XL, a new lead generation tool based on a microservice architecture that includes a component of explainable AI. (mu )XL collects and stores historical and real-time data from web sources, like Google Trends, and generates current and future leads. Leads are produced by a novel engine for hypothetical reasoning based on temporal logical rules, which can identify propositions that may hold depending on the outcomes of future events. This engine also supports additional features that are relevant for lead generation, such as user-defined predicates (allowing useful custom atomic propositions to be defined as Java functions) and negation (needed to specify and reason about leads characterized by the absence of specific properties). Our microservice architecture is designed using state-of-the-art methods and tools for API design and implementation, namely API patterns and the Jolie programming language. Thus, our development provides an additional validation of their usefulness in a new application domain (journalism). We also carry out an empirical evaluation of our tool.

Several problems involving uncertainties can be modeled with fuzzy numbers according to the type of these uncertainties. It is natural to express the solution to such a problem with fuzzy numbers. In this study, we consider the fully fuzzy transportation problem. All input parameters of the problem are expressed with fuzzy numbers given in the parametric form. We propose a new heuristic algorithm to approximate the fuzzy optimal solution. The fuzzy problem is solved by transforming it into two independent parametric problems with the proposed method. We first divide the interval [0, 1] into a sufficiently large number of equal intervals, then write a linear programming problem for each partition point and solve these problems by transforming them into transportation problems. The proposed algorithm is supported by examples.

Attack detection in cyber security systems is one of the complex tasks which require domain specific knowledge and cognitive intelligence to detect novel and unknown attacks from large scale network data. This research explores how the network operations and network security affects the detection of unknown attacks in network systems. A hash based profile matching technique is presented in this paper for attack detection. The main objective of this work is to detect unknown attacks using a profile matching approach in Hypervisors. Hypervisors are characterized by their versatile nature since they allow the utilization of available system resources. The virtual machines (VMs) in the hypervisors are not dependent on the host hardware and as a result, hypervisors are considered advantageous. In addition, hypervisors have direct access to the hardware resources such as memory, storage and processors. However, hypervisors are more susceptible to the security threats which attack each and every VM. A SHA3-512 hashing algorithm used for generating hash values in hypervisor and the proposed model is used to verify whether the profile is malicious or benign. The performance of the hashbased profile matching technique is compared with traditional hash techniques namely SHA-256 and MD5 algorithm. Results show that the proposed SHA3-512 algorithm achieves a phenomenal performance in terms of phenomenal accuracy and zero false positive rates. Simulation results also show that the computation time required by Sha3-512 algorithm is lower compared to SHA-256 and MD5 algorithms. The performance analysis validates that the hash based approach achieves reliable performance for attack detection. The effectiveness of the hashing technique was determined using three different evaluation metrics namely attack DR, FPR, and computational time. Simulation results show that the existing SHA3- 512 algorithm detection rate of 97.24% with zero false positive rate and faster computational time compared to SHA 256 and MD5 algorithms.

Deep neural networks are vulnerable to attacks, posing significant security concerns across various applications, particularly in computer vision. Adversarial training has demonstrated effectiveness in improving the robustness of deep learning models by incorporating perturbations into the input space during training. Recently, adversarial training has been successfully applied to deep recommender systems. In these systems, user and item embeddings are perturbed through a minimax game, with constraints on perturbation directions, to enhance the model’s robustness and generalization. However, they still fail to defend against iterative attacks, which have shown an over 60% increase in effectiveness in the computer vision domain. Deep recommender systems may therefore be more susceptible to iterative attacks, which might lead to generalization failures. In this paper, we adapt iterative examples for deep recommender systems. Specifically, we propose a Deep Recommender with Iteration Directional Adversarial Training (DRIDAT) that combines attention mechanism and directional adversarial training for recommendations. Firstly, we establish a consumer-product collaborative attention to convey consumers different preferences on their interested products and the distinct preferences of different consumers on the same product they like. Secondly, we train the DRIDAT objective function using adversarial learning to minimize the impact of iterative attack. In addition, the maximum direction attack could push the embedding vector of input attacks towards instances with distinct labels. We mitigate this problem by implementing suitable constraints on the direction of the attack. Finally, we perform a series of evaluations on two prominent datasets. The findings show that our methodology outperforms all other methods for all metrics.

Various hospitals have adopted digital technologies in the healthcare sector for various healthcare-related applications. Due to the effect of the Covid-19 pandemic, digital transformation has taken place in many domains, especially in the healthcare domain; it has streamlined various healthcare activities. With the advancement in technology concept of telemedicine evolved over the years and led to personalized healthcare and drug discovery. The use of machine learning (ML) technique in healthcare enables healthcare professionals to make a more accurate and early diagnosis. Training these ML models requires a massive amount of data, including patients’ personal data, that need to be protected from unethical use. Sharing these data to train ML models may violate data privacy. A distributed ML paradigm called federated learning (FL) has allowed different medical research institutions, hospitals, and healthcare devices to train ML models without sharing raw data. This survey paper overviews existing research work on FL-related use cases and applications. This paper also discusses the state-of-the-art tools and techniques available for FL research, current shortcomings, and future challenges in using FL in healthcare.

Predictive monitoring is a subfield of process mining that aims to predict how a running case will unfold in the future. One of its main challenges is forecasting the sequence of activities that will occur from a given point in time —suffix prediction—. Most approaches to the suffix prediction problem learn to predict the suffix by learning how to predict the next activity only, while also disregarding structural information present in the process model. This paper proposes a novel architecture based on an encoder-decoder model with an attention mechanism that decouples the representation learning of the prefixes from the inference phase, predicting only the activities of the suffix. During the inference phase, this architecture is extended with a heuristic search algorithm that selects the most probable suffix according to both the structural information extracted from the process model and the information extracted from the log. Our approach has been tested using 12 public event logs against 6 different state-of-the-art proposals, showing that it significantly outperforms these proposals.

One of the preconditions for efficient cloud computing services is the continuous availability of services to clients. However, there are various reasons for temporary service unavailability due to routine maintenance, load balancing, cyber-attacks, power management, fault tolerance, emergency incident response, and resource usage. Live Virtual Machine Migration (LVM) is an option to address service unavailability by moving virtual machines between hosts without disrupting running services. Pre-copy memory migration is a common LVM approach used in cloud systems, but it faces challenges due to the high rate of frequently updated memory pages known as dirty pages. Transferring these dirty pages during pre-copy migration prolongs the overall migration time. If there are large numbers of remaining memory pages after a predefined iteration of page transfer, the stop-and-copy phase is initiated, which significantly increases downtime and negatively impacts service availability. To mitigate this issue, we introduce a prediction-based approach that optimizes the migration process by dynamically halting the iteration phase when the predicted downtime falls below a predefined threshold. Our proposed machine learning method was rigorously evaluated through experiments conducted on a dedicated testbed using KVM/QEMU technology, involving different VM sizes and memory-intensive workloads. A comparative analysis against proposed pre-copy methods and default migration approach reveals a remarkable improvement, with an average 64.91% reduction in downtime for different RAM configurations in high-write-intensive workloads, along with an average reduction in total migration time of approximately 85.81%. These findings underscore the practical advantages of our method in reducing service disruptions during live virtual machine migration in cloud systems.