No abstract

No abstract

No abstract

Writing a survey paper on one research topic usually needs to cover the salient content from numerous related papers, which can be modeled as a multi-document summarization (MDS) task. Existing MDS datasets usually focus on producing the structureless summary covering a few input documents. Meanwhile, previous structured summary generation works focus on summarizing a single document into a multi-section summary. These existing datasets and methods cannot meet the requirements of summarizing numerous academic papers into a structured summary. To deal with the scarcity of available data, we propose BigSurvey, the first large-scale dataset for generating comprehensive summaries of numerous academic papers on each topic. We collect target summaries from more than seven thousand survey papers and utilize their 430 thousand reference papers’ abstracts as input documents. To organize the diverse content from dozens of input documents and ensure the efficiency of processing long text sequences, we propose a summarization method named category-based alignment and sparse transformer (CAST). The experimental results show that our CAST method outperforms various advanced summarization methods.

Graph neural networks (GNNs) have achieved great success on various graph tasks. However, recent studies have revealed that GNNs are vulnerable to adversarial attacks, including topology modifications and feature perturbations. Regardless of the fruitful progress, existing attackers require node labels and GNN parameters to optimize a bi-level problem, or cannot cover both topology modifications and feature perturbations, which are not practical, efficient, or effective. In this paper, we propose a black-box attacker PEEGA, which is restricted to access node features and graph topology for practicability. Specifically, we propose to measure the negative impact of various adversarial attacks from the perspective of node representations, thereby we formulate a single-level problem that can be efficiently solved. Furthermore, we observe that existing attackers tend to blur the context of nodes through adding edges between nodes with different labels. As a result, GNNs are unable to recognize nodes. Based on this observation, we propose a GNN defender GNAT, which incorporates three augmented graphs, i.e., a topology graph, a feature graph, and an ego graph, to make the context of nodes more distinguishable. Extensive experiments on three real-world datasets demonstrate the effectiveness and efficiency of our proposed attacker, despite the fact that we do not access node labels and GNN parameters. Moreover, the effectiveness and efficiency of our proposed defender are also validated by substantial experiments.

The FDDI Standard speci"es a 100 Mbits/s "bre-optic token ring network that has been implemented and widely installed. The FDDI medium access control (MAC) protocol, however, has the inherent de"ciency that, at most, one half of the bandwidth of a FDDI ring can be used to transmit synchronous messages. In a recent paper, a modi"cation to the FDDI MAC protocol, called FDDI-M, was proposed to overcome this de"ciency. It has been shown using simulation that FDDI-M doubles a ring's ability of supporting synchronous traf"c while at the same time it achieves a higher throughput for asynchronous traf"c than standard FDDI. In this paper we present an analytical study of the timing properties of the FDDI-M protocol. The results presented in this paper complement those of the original paper. Using the worst-case achievable utilization (WCAU) as the performance metric, we evaluate the performance of various synchronous bandwidth allocation (SBA) schemes in guaranteeing synchronous message deadlines. It is found that, in comparison with FDDI, the FDDI-M protocol results in a higher WCAU for the normalized proportional SBA scheme. However, for the local SBA schemes studied, the WCAU values remain at zero.

Human errors, e.g. surgeon's misoperation, have been recognised as a critical cause to the large amount of medical accidents in hospital. Recently, the concept of Medical Cyber-Physical Systems (MCPS) has been proposed to enable automatic medical device coordination for patient protection. However, MCPS have limited capabilities to detect human errors because of only integrating medical devices, and thus, often result in late device coordination when patients are found to have already developed significant adverse physiological reactions. In this paper, we propose to build context-aware MCPS to avoid such risky situations. We leverage various nonmedical devices to capture implicit contextual information when human users are interacting with MCPS. By using these contexts, we significantly raise the system's awareness to human errors, and thus, allow it to take proper actions as early as possible to avoid the potential accidents. A major challenge in designing such systems, however, is how to deal with context uncertainty without sacrificing patient safety. Contexts are uncertain in nature, but false context detection can trigger unnecessary actions harmful to patient. To address this issue, we develop a novel scheme called 'context-assessment-action', where medical knowledge is utilised to assess all context-triggered actions and prohibit the risky ones. To our knowledge, our approach is the first to enable context-awareness for safety-critical systems. Finally, we apply this approach and conduct a case study on patient-controlled analgesia. Experimental results demonstrate the effectiveness of our approach and the great promise of context-aware MCPS for patient safety improvement.

No abstract