Yulei Wu’s research while affiliated with University of Bristol and other places

... 2) Epistemic logic: Epistemic logic is a branch of formal reasoning dealing with the inference, transfer, and update of knowledge among multiple agents [32], [33]. When knowledge evolves over time and successive interactions, this is referred to as DEL, and finds applications in social networks and cryptography [34]. ...

Reference:

Distributed Optimization of Age of Incorrect Information with Dynamic Epistemic Logic

... STGAFormer [12] introduced adaptive adjacency matrices to capture these hidden spatial correlations, but this learns graph structures that will remain unchanged from period to period, ignoring the dynamics of the graph structure. AdpSTGCN [13] used an attention mechanism to fexibly generate multiview feature maps, obtaining local dynamic spatial features through a digitally driven approach, yet ignoring the overall static spatial hidden features such as urban POIs. COGCN [14] used OAG to extract global spatial features of the road and ODPG to extract local spatial features of the road and combined them using a comparative learning algorithm. ...

Reference:

Optimal Graph Information Fused Graph Attention Network for Traffic Flow Forecasting

... The traditional methods of evaluating decisions, such as cost-benefit analysis, do not fit the criteria for measuring ethical dilemmas in AI. Further, in instances like an autonomous vehicle, ethical decision-making should involve moral dilemmas inside and outside the vehicle (Huang et al., 2024). Principles alone cannot guarantee ethical AI (B. ...

Reference:

Ethical AI and Decision-Making in Management Leadership

... These hurdles can resist timely access to critical information, disrupt online learning activities, and overburden the mainstream network 18 . Delays in accessing learning materials or receiving feedback may restrict the learning process and decrease student engagement, while network congestion and performance issues can disrupt lectures, discussions, and collaborative activities, impacting effective communication and interaction 19 . Prolonged disruptions could lead to decreased productivity and motivation among students, ultimately affecting their academic performance and overall learning outcomes. ...

Reference:

Regional computing approach for educational big data

... In EC systems, the resource allocation in current works is often conducted for response latency minimization or energy consumption saving. Various approaches, including game-theoretic methods [7], heuristic algorithms [8], or deep reinforcement learning techniques [9], have been employed to tackle these resource allocation challenges. ...

Reference:

Joint Optimization of Service Caching Task Offloading and Resource Allocation in Cloud-Edge Cooperative Network

... However, others include a variety of human evaluations to verify the results of the automated evaluation. Common assessments include the validity of the adversarial perturbation, the classification accuracy of the original task by humans (Jin et al. 2020;Alzantot et al. 2018;Garg and Ramakrishnan 2020;Li et al. 2020); the similarity of the adversarial text to the original (Jin et al. 2020;Alzantot et al. 2018;Li et al. 2023Li et al. , 2021Li et al. , 2020; and grammatical correctness (Jin et al. 2020;Li et al. 2023Li et al. , 2021Li et al. , 2020. ...

Reference:

Graded Suspiciousness of Adversarial Texts to Humans

... Other techniques find idle GPUs and then place tasks [31], or leverage the migration capacities of containers to access GPU servers in remote locations. GPU allocation is based on the location of the user in an edgecomputing environment [32,33]. ...

Reference:

GPU Usage Time-Based Ordering Management Technique for Tasks Execution to Prevent Running Failures of GPU Tasks in Container Environments

... The multi-agent task allocation problem involves distributing tasks among multiple agents to optimize resource utilization and minimize operational costs [1,2]. Effective coordination among agents is essential to enhance performance across a range of applications [3,4]. Effective task allocation is crucial in these contexts, as it directly impacts the system's overall performance, efficiency, and success [5]. ...

Reference:

Efficient Task Allocation in Multi-Agent Systems Using Reinforcement Learning and Genetic Algorithm

... Although meta-learning can be useful, its costs must first be evaluated. In some domains, data is limited but annotation of even limited data requires experts and results in high costs [129]. Meta-learning can also be difficult to scale due to the high computational and memory costs, training instability, and lack of efficient distributed training support [130,131]. ...

Reference:

A Comprehensive Survey of Challenges and Opportunities of Few-Shot Learning Across Multiple Domains

... Contrastive learning (CL) can alleviate the problem of data scarcity, especially by improving data quality and utilizing limited labeled data, and has gradually been applied to rumor detection (Lin et al., 2022;Xu et al., 2023;Gao et al., 2023;Cui and Jia, 2024). In addition, multi-task learning (MTL) is a strategy that improves the generalization ability of the model by leveraging the shared knowledge between multiple interrelated tasks (Caruana, 1997), which has also been widely explored and used in the field of rumor detection (Zhang et al., 2021;Zhang and Gao, 2024). ...

Reference:

Deciphering Rumors: A Multi-Task Learning Approach with Intent-aware Hierarchical Contrastive Learning