Large language models(LLMs)have demonstrated impressive capabilities across a wide range of reasoning tasks.To address the lack of faithfulness in prompt-based reasoning methods such as Chain-of-Thought(CoT),researchers have explored neuro-symbol-ic approaches to bolster LLM logical reasoning capabilities.However,existing methods still face challenges with information loss dur-ing.To overcome these limitations,this paper introduce Iterative Feedback-Driven Neuro-Symbolic(IFDNS),a novel prompt-based method that employs a multi-round feedback mechanism to address LLM limitations in handling complex logical relationships.IFDNS utilizes iterative feedback during the logic extraction phase to accurately extract causal relationship statements and translate them into propositional and logical implication expressions,effectively mitigating information loss issues.Furthermore,IFDNS is orthogonal to existing prompt methods,allowing for seamless integration with various prompting approaches.Empirical evaluations across six data-sets demonstrate the effectiveness of IFDNS in significantly improving the performance of CoT and Chain-of-Thought with Self-Con-sistency(CoT-SC).
In the field of medical image analysis, the scarcity of Chinese chest X-ray report datasets has hindered the development of technology for generating Chinese chest X-ray reports. On one hand, the construction of a Chinese chest X-ray report dataset is limited by the time-consuming and costly process of accurate expert disease annotation. On the other hand, a single natural language generation metric is commonly used to evaluate the similarity between generated and ground-truth reports, while the clinical accuracy and effectiveness of the generated reports rely on an accurate disease labeler (classifier). To address the issues, this study proposes a disease labeler tailored for the generation of Chinese chest X-ray reports. This labeler leverages a dual BERT architecture to handle diagnostic reports and clinical information separately and constructs a hierarchical label learning algorithm based on the affiliation between diseases and body parts to enhance text classification performance. Utilizing this disease labeler, a Chinese chest X-ray report dataset comprising 51,262 report samples was established. Finally, experiments and analyses were conducted on a subset of expert-annotated Chinese chest X-ray reports, validating the effectiveness of the proposed disease labeler.
With the expansion of network scales, the B/S architecture of monolithic applications is gradually being replaced by microservices. The unbundling of services has led to exponential growth in the size of APIs. When handling massive microservice requests, the commercial NIC shows limitations in three aspects: deterministic, programmability, and data copy. To ensure that each microservice node handles requests efficiently, flexibly, and precisely, this paper proposes a programmable deterministic multi-queue FPGA Accelerator. The Accelerator relies on the instantiated 1000 queues and the queue management unit to extend the rule-based RSS algorithm for the serverless-friendly programmability of packet distribution. A PTP hardware clock is added to collaborate with the queue management unit to control the deterministic delivery. To improve the sending and receiving efficiency of network node data, a driver adapted to the FPGA accelerator is designed to realize zero-copy. Experiments conducted on a 100 Gbps FPGA show that the Accelerator can support the multi-queue transmission with various packet sizes, define the forwarding behavior, and almost approach the line rate on an 8-core FPGA device. In addition, it can forward packets with low latency close to that of the current state-of-the-art ovs-DPDK. This Accelerator overcomes, to some extent, the limitations of commercial NICs when oriented to microservice architectures.
In knowledge discovery and data mining, removing redundant and irrelevant data attributes is crucial. Traditional algorithms, however, struggle with efficiency in high-dimensional big data contexts. To solve this problem, this paper proposes a novel global search attribute reduction method namely RHC-IGWO (Rough Hypercuboid Structure via Improved Grey Wolf Optimizer) by integrating the rough hypercuboid method and Improved Grey Wolf Optimizer (IGWO) with the pheromone mechanism. The algorithm is embedded into the Apache Spark parallel computing framework (Parallel computing Rough Hypercuboid Structure via Improved Grey Wolf Optimizer, PcRHC-IGWO) to accelerate and simplify the attribute reduction process. The algorithm divides the decision table into several independent blocks, introduces the pheromone mechanism to simulate the wolf pack behavior, and uses the IGWO for global search, which is conducive to efficient local search and global information sharing between individuals. The position of the individual is initialized by calculating the relevance between the attributes, and the pheromone value is dynamically updated according to the reduction quality. This allows automatically giving more search focus to more promising attribute regions. Experiments with public and real datasets demonstrate the RHC-IGWO algorithm's significant speedup and its efficacy in maintaining or enhancing classification accuracy. Particularly noteworthy is its performance on schizophrenia datasets, where the proposed method achieves outstanding classification accuracies of 86.2%, 88.89%, and 92.86% across various classifiers. These results not only demonstrate its effectiveness but also underline its potential in advanced data analysis scenarios. Additionally, on some large-scale datasets, the time required for processing has been reduced by 85.71%, showcasing the algorithm's efficiency in handling complex and voluminous data.
Graph representation learning aims to learn embedding representations of nodes in graph-structured data through supervised or unsupervised methods. However, for unlabeled or sparsely labeled graph datasets, conventional supervised approaches fail to perform adequately, and existing unsupervised methods have difficulty achieving high accuracy on multiple downstream tasks such as node classification and link prediction at lower embedding dimensions. To address this issue, we propose a novel self-supervised graph representation learning approach called GAN-based Message Passing Graph Representation Learning. This approach significantly improves the accuracy of node embeddings without relying on label information. Specifically, GMP-GL has developed a generator based on neighbor-weighted random walk to obtain positive and negative node sample pairs, as well as a discriminator that distinguishes whether a real connection exists between node pairs. Through adversarial learning, GMP-GL automatically learns node embeddings. Furthermore, GMP-GL optimizes the training outcomes of the generative adversarial model through self-supervised embedding aggregation learning, thus introducing rich structural information from the graph for the subsequent rounds of generative adversarial training. In addition, GMP-GL employs multiple autoencoders to generate and filter original embeddings, ensuring that the pre-embeddings preserve a significant amount of original feature information for formal learning. The study selects six real-world datasets and conducts comparative experiments with ten SOTA baseline methods. The experimental results demonstrate that the proposed model not only surpasses previous unsupervised methods but also matches or even exceeds the performance of supervised methods. This indicates the great potential of the proposed approach for representation learning tasks on unlabeled graph datasets and its effective applicability to real-world problems.
Key expansion plays an important role in symmetric block cipher. To generate desired number of round keys that satisfy irreversibility, parallelism and independence, we first constructed a 2D memristor-based discrete hyper chaotic map (2D-MDHCM), through adding a memristor to the 2D discrete hyper chaotic map (2D-DHCM), whose state point has ergodicity in phase space and good randomness in a sufficiently large parameter range. Took advantage of the irreversibility of the chaotic iteration, we designed an irreversible key expansion algorithm, which can generate the round keys in parallel with specified round and flexible length, what’s more, each round key can be generated by the initial key independently. Experimental and analysis results demonstrated that the improved key expansion algorithm satisfied irreversibility, parallelism and independence, which enhanced the security of initial key greatly.
Load prediction is an essential technique to improve edge system performance by proactively configuring and allocating system resources. Traditional load prediction methods obtain high prediction when handling loads exhibiting cyclical trend behavior, but they are unable to capturing highly-variable loads in edge computing environments. Existing studies fit prediction models via independent time series and output single-point real-value predictions. However, in practical edge scenarios, it is more valuable to obtain application value by utilizing the probability distribution of future loads rather than directly predicting specific values. To solve these problems, we propose an Edge Load Prediction method empowered by Deep Auto-regressive Recurrent networks (ELP-DAR). The ELP-DAR uses the time-series data of edge loads to train deep auto-regressive recurrent networks, which integrate Long Short-Term Memory (LSTM) into the S2S framework to calculate the parameters of the probability distribution at the next time-point. Therefore, the ELP-DAR can efficiently extract the essential representations of edge loads and learn their complex patterns, and the probability distribution for highly-variable edge loads can be accurately predicted. Extensive simulation experiments are conducted to validate the effectiveness of the proposed ELP-DAR method based on real-world edge load datasets. The results show that the ELP-DAR achieves higher prediction accuracy than other benchmark methods with different prediction lengths.
With the development of blockchain technology, consensus algorithm, as its core technology, has also become a research hotspot of many scholars.A good consensus algorithm can effectively solve the problems of blockchain, such as security, scalability, energy cost and performance efficiency.This paper selects Raft consensus algorithm as the research object.Aiming at the election delay caused by split voting in Raft algorithm leader election, an efficient consensus algorithm based on PoW-RPFT consensus algorithm is proposed.Firstly, the PoW consensus algorithm is used to successfully select the efficient deputy leader node; Secondly, each node is given a waiting time, and the waiting time is adjusted according to the node behavior; Finally, the waiting time election model is introduced, combined with PoW consensus algorithm to optimize the vote election of Raft algorithm, so as to quickly elect an efficient leader node.The experimental results show that the new RPFT algorithm improves the election speed by about 75% compared with Raft algorithm; In terms of consensus efficiency, the consensus efficiency of RPFT algorithm is improved by 40%.
针对室内环境下Wi-Fi信号波动及在线定位阶段移动设备异质性而导致指纹定位算法精度不高的问题,本文考虑了用户移动轨迹上一系列接收信号强度指示(Received Signal Strength Indicator,RSSI)指纹之间的相关性,并提出了一种基于深度神经网络的室内定位算法.该算法在离线阶段,针对部分手机难以采集大量RSSI指纹的问题,通过数据增强扩充指纹库,减少指纹采集工作量;然后使用轨迹生成算法生成大量轨迹数据,训练基于卷积神经网络和循环神经网络的定位模型.在线阶段将RSSI指纹转换为差分矩阵以缓解设备异质性问题,并结合指纹的隐含特征以及模型上一时刻预测的位置进行定位.在不同移动终端设备上的实验结果表明,该算法可以有效缓解设备异质性的影响,提高定位精度.
现有基于深度学习的单棵树木检测方法一般是对同个树木对象产生重复的候选框,然后通过非极大值抑制方法进行过滤,以确定每个对象最终的边框.这种方法对于孤立的树木对象能够起到很好的作用,但对密集重叠的树木场景中的单棵树木检测结果并不理想.本文提出一种基于迭代框架的密集场景单棵树木检测方法.该方法采用迭代框架统合两轮训练过程,首先在第1轮训练中提取边框,然后在第2轮训练中引入迭代边框,对候选边框进行筛除后再执行非极大值抑制过程,最后将非极大值抑制结果和迭代边框进行合并整合,这种方法将重叠边框分配到两轮训练中分别处理,减少了候选框之间的干扰;此外,该方法在两轮训练中可以采用不同的网络,本文通过实验得到最为适配的组合.为验证本文方法的有效性,本文采集和标注了密集重叠场景的树木数据集,并在该数据集上进行了实验和结果分析,实验结果表明本文提出的方法与现有方法相比,树木检测的准确率和召回率分别提升了 2.4％和2.7％,能够有效提升密集场景树木检测的准确率.
基于网络功能虚拟化(Network Function Virtualization,NFV)软硬件解耦的网络架构,运营商通过部署多个端到端逻辑网络为垂直行业提供多样化服务.然而,面对时延敏感型服务,需要将部分虚拟网络功能(Virtual Network Function,VNF)下沉到网络边缘.基于此,多接入边缘计算(Multi-Access Edge Computing,MEC)技术应运而生.但是,与核心云相比,边缘云的资源容量和计算能力有限,因此,在部署网络服务时既要考虑云边协同下的资源分配问题,还要考虑如何进行流量调度以提升边缘云的资源利用率并解决底层网络中的负载均衡问题.鉴于此,本文提出了一种基于动态流量拆分的VNF部署和流量调度方法.通过动态流量拆分将流量请求分配到多条路径和多个节点,从而使全局流量分布的更加均衡,根据流量拆分结果进行VNF部署,进而减少发生网络拥塞和流量请求超时的概率.其中,每条流量请求的拆分条数与拆分比例可以根据底层网络的资源状态和流量的特性灵活调整.因此,整个过程包括有效的流量拆分策略、VNF部署策略和流量调度策略.本文将云边协同下的虚拟网络功能部署和流量调度问题描述为混合整数线性规划(Mixed Integer Linear Programming,MILP)问题,并提出了一种基于禁忌搜索算法和遗传算法的联合优化(TSGA)算法.仿真结果表明,TSGA算法相比于TS算法和RB算法可以分别提高7.7％和12.9％的流量请求接受率,并分别减少5.4％和7.8％的流量平均开销.
数据竞争是并行程序中最常见的问题,因为其在程序中存在着随机性和难以重现的问题,所以利用动态方法检测并不可靠.本文介绍了一种针对以OpenMP实现程序并行的数据竞争故障的静态检测方法.首先根据基于OpenMP的Fortran并行程序的语法和语义特性,构建并行程序的抽象语法树,并对控制流图进行扩展引入并行控制流图;然后将数据竞争问题抽象为一种故障模型,利用有限状态机来形式化的描述此类故障模型;最后对并行区域的潜在赋值操作进行并行数据流分析,在程序控制流图节点上进行故障状态机的状态转化来实现数据竞争的检测.该方法已在自研的缺陷检测系统(DTS)中应用,并通过DataRaceBench的实验数据证明本文方法可以有效的检测出数据竞争问题.
微服务架构由于其高效灵活的独特优势近年来已逐渐成为一种代替单体架构的新选择,在企业实践中,单体架构系统向微服务架构系统的迁移过程通常是由人工完成的,这种方式往往不确定性高,主观性过强.为了解决这些问题,提出了 一种组合优化模型驱动的微服务拆分方法,该方法将微服务拆分问题建模成为一种组合优化问题,基于微服务的高内聚低耦合原则和业务一致性原则生成组合优化的目标函数,然后通过人工鱼群算法得到最优的微服务拆分方案.实验表明,该方法能有效提高微服务拆分的质量,得到的拆分结果更为合理可靠.
质量估计的目的是在没有参考译文的情况下衡量翻译内容的质量,这对于需要高质量翻译任务中的机器翻译系统至关重要.针对有监督的翻译质量估计中普遍存在的缺乏标记训练数据和模型框架中两阶段学习目标差异的问题,以及无监督的质量估计中存在的因为任务目标模糊、特征学习不完全而导致最终结果远不如有监督模型的问题,本文提出了将无监督质量估计中的估计指标当作特征融入有监督的质量估计模型中的方法,以此来互相弥补两种模型之间存在的缺点.在WMT2020的高资源对和低资源对上的实验结果证实,相对于基线的有监督和无监督模型,两者结合的方法能够更好的提高翻译质量估计的准确性,与人工评分的皮尔逊相关系数都有所提升.
超图直推学习模型是机器学习领域研究热点.超图模型的性能取决于构造的超图结构及其Laplacian矩阵的质量.现有超图模型基于单一超图结构,信息表达能力有限.本文提出超图结构扩张法,将异构超图的关联矩阵和权重矩阵拼接,融合更多的顶点间全局高阶信息,增加Markov随机游走的扩散范围.但这会导致矩阵维度高,计算开销大.因此进一步提出多通道Laplacian矩阵融合法,用多个通道计算异构超图结构各自的Laplacian矩阵,再加权累加.在4个数据集上的实验表明,两种方法都能提高超图直推学习模型的分类性能,且Laplacian矩阵融合法比结构扩张法平均节约40％左右时间成本,Fl指标最高提升 8.4％.
开源软件发展迅猛,对于开源项目来说,开源许可证的选择已然成为最为关键性的问题,考虑项目中所用开源组件的许可证兼容问题,稍有不慎就会产生多种多样的风险问题.怎样检测项目所含许可证,如何选择项目的开源许可证,项目内部许可证是否冲突.旨在解决这些问题的基础上,提出一种基于BERT模型实现文本相似度检测,辅以URL检测和AC自动机进行关键字的抓取的综合检测方法,通过多样化检测方式提高检测准确度,最后再采用有向图算法进行许可证推荐.对于许可证的检测方法,通过实验结果表明该方法相比于传统方法有更好的检测效果.对于许可证的推荐方法,实验表明该方法可以有效地检测出兼容性推荐结果.
随着我国人口老龄化程度持续加深,老年人对于社区居家护理服务需求量不断增加,老年人的满意度保障问题亟待解决,合理调度护理人员以提高社区居家养老服务质量或将有效解决该问题.针对社区居家养老护理人员调度问题,建立基于模糊时间窗的老年人满意度函数,构建最小化护理人员数量和最大化老年人满意度的多目标优化模型.在标准粒子群算法的基础上,使用一种新的惯性因子取值方法,并引入大邻域搜索算子以提高算法局部寻优能力,设计出混合粒子群算法进行求解.最后,进行数值实验和算法对比,结果表明,提出的混合粒子群算法能够有效解决老年人聚集分布情况下的护理人员调度问题.
事件检测是信息抽取的关键子任务,目的是识别文本中特定类型的事件实例.图注意力网络在图结构上使用自注意力机制,能实现高质量的事件检测,但现有图注意力网络存在忽略依存标签信息、上下文信息获取模型复杂等问题,导致模型准确率下降、计算资源开销较大.本文提出依存边信息嵌入的图注意力网络模型(EIEGAT),设计依存边信息嵌入模块,将依存标签信息嵌入到图的邻接矩阵中,使模型在构造图时同时考虑节点和依赖边的表示,提升事件检测准确率.使用结构和计算更简单的门控循环单元捕获单词的上下文信息,在维持性能的同时简化网络结构、节省内存空间.实验表明,EIEGAT有效提高事件检测的总体性能,在ACE2005英文语料集上事件识别与事件分类的F1值分别提高5％与0.9％.
为了解人工智能(artificial intelligence,AI)的研究现状和研究热点,以近10年Scopus和Web of Science数据库中的期刊与会议论文为研究对象,运用文献计量学方法,分析国内外AI学术研究现状,包括全球AI学术研究发展态势、发表论文国家/区域分布、AI研究机构分布、AI学者分布.在此基础上,进行AI研究热点分析,包括AI学科的知识结构和AI研究的关键词和研究主题,为人工智能相关的科学研究和管理决策提供参考.研究结果显示:1)全球人工智能领域研究成果总体上呈现急速增长态势,但仍然处于快速发展阶段,具有巨大发展空间和广阔发展前景;2)美国、中国及其研究机构在人工智能研究的文献数量方面占有绝对优势,但中国与全球主要国家与机构相比科研绩效较低;3)人工智能与众多学科的交叉融合已经成为主流,深度学习是人工智能的第一大研究方向,研究热点向着物联网和大数据转变,图像识别、工业4.0、意见挖掘、推荐系统、人机互动、脑机接口、轨迹跟踪、自动驾驶、故障检测、网络安全、交通流预测、机器翻译等方面是与人工智能相关的研究热点前沿;4)中国作为世界第一发文量大国,人工智能研究应当进一步重视提升研究质量,以迎接国际科技竞争的挑战并抢占全球新一轮科技革命和产业变革发展制高点.
开源指令集RISC-V为物联网和嵌入式领域的处理器提供了强大的动力,本文针对一些具有高性能、小面积、低功耗需求的场景,设计了一种基于RISC-V指令集架构的高性能嵌入式处理器核.处理器核的代号为FRV232,采用单取指,单发射,乱序执行技术,支持RV32I基础指令集和M扩展指令集,以较低的面积实现了较高的性能.本文开发了专门用于验证FRV232核心的功能模型,功能验证阶段使用验证软件Modelsim和功能模型对处理器核心进行了完整的验证,并利用该处理器核心在FPGA上实现了基础的原型系统,使用Vivado统计了该处理器核所需的芯片面积.经过测试,FRV232在FPGA上能够稳定运行在100MHz,在该主频下,Dhrystone的性能跑分可以达到1.73DMPS/MHz.
