On localized control in QoS routing

In this note, we study several issues in the design of localized quality-of-service (QoS) routing schemes that make routing decisions based on locally collected QoS state information (i.e., there is no network-wide information exchange among routers). In particular, we investigate the granularity of local QoS state information and its impact on the design of localized QoS routing schemes from a theoretical perspective. We develop two theoretical models for studying localized proportional routing: one using the link-level information and the other using path-level information. We compare the performance of these localized proportional routing models with that of a global optimal proportional model that has knowledge of the global network QoS state. We demonstrate that using only coarser-grain path-level information it is possible to obtain near-optimal proportions. We then discuss the issues involved in implementation of localized proportional routing and present some practical schemes that are simple and easy to implement     

Distance estimation and object location via rings of neighbors

We consider four problems on distance estimation and object location which share the common flavor of capturing global information via informative node labels: low-stretch routing schemes [48], distance labeling [25], searchable small worlds [31], and triangulation-based distance estimation [34]. Focusing on metrics of low doubling dimension, we approach these problems with a common technique called rings of neighbors, which refers to a sparse distributed data structure that underlies all our constructions. Apart from improving the previously known bounds for these problems, our contributions include extending Kleinberg’s small world model to doubling metrics, and a short proof of the main result in Chan et al. [15]. Doubling dimension is a notion of dimensionality for general metrics that has recently become a useful algorithmic concept in the theoretical computer science literature. This work was done when A. Slivkins was a graduate student at Cornell University and was supported by the Packard Fellowship of Jon Kleinberg. Preliminary version of this paper has appeared in 24th Annual ACM SIGACT-SIGOPS Symposium on Principles of Distributed Computing (PODC), 2005.     

融合全局和局部特征的下一个兴趣点推荐方法

随着海量移动数据的积累,下一个兴趣点推荐已成为基于位置的社交网络中的一项重要任务.目前,主流方法倾向于从用户近期的签到序列中捕捉局部动态偏好,但忽略了历史移动数据蕴含的全局静态信息,从而阻碍了对用户偏好的进一步挖掘,影响了推荐的准确性.为此,提出一种基于全局和局部特征融合的下一个兴趣点推荐方法.该方法利用签到序列中的顺序依赖和全局静态信息中用户与兴趣点之间、连续签到之间隐藏的关联关系建模用户移动行为.首先,引入两类全局静态信息,即User-POI关联路径和POI-POI关联路径,学习用户的全局静态偏好和连续签到之间的全局依赖关系.具体地,利用交互数据以及地理信息构建异构信息网络,设计关联关系表示学习方法,利用相关度引导的路径采样策略以及层级注意力机制获取全局静态特征.然后,基于两类全局静态特征更新签到序列中的兴趣点表示,并采用位置与时间间隔感知的自注意力机制来捕捉用户签到序列中签到之间的局部顺序依赖,进而评估用户访问兴趣点概率,实现下一个兴趣点推荐.最后,在两个真实数据集上进行了实验比较与分析,验证了所提方法能够有效提升下一个兴趣点推荐的准确性.此外,案例分析表明,建模显式路径有助于提...     

Labeling schemes for weighted dynamic trees

A Distance labeling scheme is a type of localized network representation in which short labels are assigned to the vertices, allowing one to infer the distance between any two vertices directly from their labels, without using any additional information sources. As most applications for network representations in general, and distance labeling schemes in particular, concern large and dynamically changing networks, it is of interest to focus on distributed dynamic labeling schemes. The paper considers dynamic weighted trees where the vertices of the trees are fixed but the (positive integral) weights of the edges may change. The two models considered are the edge-dynamic model, where from time to time some edge changes its weight by a fixed quanta, and the increasing-dynamic model in which edge weights can only grow. The paper presents distributed approximate distance labeling schemes for the two dynamic models, which are efficient in terms of the required label size and communication complexity involved in updating the labels following the weight changes.     

一种基于深度学习的多尺度深度网络的场景标注算法

针对场景标注中如何产生良好的内部视觉信息表达和有效利用上下文语义信息两个至关重要的问题,提出一种基于深度学习的多尺度深度网络监督模型。与传统多尺度方法不同,模型主要由两个深度卷积网络组成:首先网络考虑全局信息,提取大尺度图像低层特征;其次网络利用图像局部信息,结合低层特征获取一组稠密的、完备的图像特征,有效地捕获图像像素的纹理特征、颜色特征和上下文信息。对比许多经典方法,该算法不依赖图像分割技术和人工制作特征,在Stanford Background Dataset取得了很好的效果。     

Sparse communication networks and efficient routing in the plane

Traditional approaches to network design separate the issues of designing the network itself and designing its management and control subsystems. This paper proposes an approach termed routing-oriented network design, which is based on designing the network topology and its routing scheme together, attempting to optimize some of the relevant parameters of both simultaneously. This approach is explored by considering the design of communication networks supporting efficient routing in the special case of points located in the Euclidean plane. The desirable network parameters considered include low degree and small number of communication links. The desirable routing parameters considered include small routing tables, small number of hops and low routing stretch. Two rather different schemes are presented, one based on direct navigation in the plane and the other based on efficient hierarchical tree covers. On a collection of n sites with diameter D, these methods yield networks with a total of communication links and some bounds on the degree, coupled with routing schemes with constant routing stretch, memory bits per vertex and routes with at most or hops. Received: October 2000 / Accepted: May 2001     

Ad hoc网络的基于债务关系的合作激励机制

合作的激励机制对拓展Ad hoc网络的应用具有决定性作用,然而在资源有限的网络条件下实施激励机制对协议开销控制的要求很高。依托文件共享应用,提出了基于债务关系的文件共享激励机制SDE。由于只涉及直接交互信息,避免了针对全局声誉方案的欺骗与攻击行为及虚拟货币方案所存在的货币造假问题,易于分布式实现。通过引入节点间的重复博弈,节点自发维护债务的有效性。提出了债务网络上的路由算法M-PGA,该算法采用局部路由思想,可以匹配债务网络动态性,而基于最优路径的路由选择可提高路由的成功率。     

Supervised neighborhood regularized collaborative representation for face recognition

How to represent a test sample is very crucial for linear representation based classification. The famous sparse representation focuses on employing linear combination of small samples to represent the query sample. However, the local structure and label information of data are neglected. Recently, locality-constrained collaborative representation (LCCR) has been proposed and integrates a kind of locality-constrained term into the collaborative representation scheme. For each test sample, LCCR mainly considers its neighbors to deal with noise and LCCR is robust to various corruptions. However, the nearby samples may not belong to the same class. To deal with this situation, in this paper, we not only utilize the positive effect of neighbors, but also consider the side effect of neighbors. A novel supervised neighborhood regularized collaborative representation (SNRCR) is proposed, which employs the local structure of data and the label information of neighbors to improve the discriminative capability of the coding vector. The objective function of SNRCR obtains the global optimal solution. Many experiments are conducted over six face data sets and the results show that SNRCR outperforms other algorithms in most case, especially when the size of training data is relatively small. We also analyze the differences between SNRCR and LCCR.     

Shape description by medial surface construction

The medial surface is a skeletal abstraction of a solid that provides useful shape information, which compliments existing model representation schemes. The medial surface and its associated topological entities are defined, and an algorithm for computing the medial surface of a large class of B-rep solids is then presented. The algorithm is based on the domain Delaunay triangulation of a relatively sparse distribution of points, which are generated on the boundary of the object. This strategy is adaptive in that the boundary point set is refined to guarantee a correct topological representation of the medial surface     

Fault-tolerant routing in hypercube multicomputers using local safety information

This paper studies fault-tolerant routing for injured hypercubes using local safety information. It is shown that a minimum feasible path is always available if the spanning subcube that contains both source and destination is safe. The safety information outside the spanning subcube is applied only when derouting is needed. A routing scheme based on local safety information is proposed and the extra cost to obtain local safety information is comparable to the one based on global safety information. The proposed algorithm guarantees to find a minimum feasible path if the spanning subcube is contained in a maximal safe subcube and the source is locally safe in the maximal safe subcube. A new technique to set up a partial path is proposed based on local safety information when the above conditions are not met. Sufficient simulation results are provided to demonstrate the effectiveness of the method by comparing with the previous methods.     

Utilizing various sparsity measures for enhancing accuracy of collaborative recommender systems based on local and global similarities

Collaborative filtering is a popular recommendation technique, which suggests items to users by exploiting past user-item interactions involving affinities between pairs of users or items. In spite of their huge success they suffer from a range of problems, the most fundamental being that of data sparsity. When the rating matrix is sparse, local similarity measures yield a poor neighborhood set thus affecting the recommendation quality. In such cases global similarity measures can be used to enrich the neighborhood set by considering transitive relationships among users even in the absence of any common experiences. In this work we propose a recommender system framework utilizing both local and global similarities, taking into account not only the overall sparsity in the rating data, but also sparsity at the user-item level. Several schemes are proposed, based on various sparsity measures pertaining to the active user, for the estimation of the parameter α, that allows the variation of the importance given to the global user similarity with regards to local user similarity. Furthermore, we propose an automatic scheme for weighting the various sparsity measures, through evolutionary approach, to obtain a unified measure of sparsity (UMS). In order to take maximum possible advantage of the various sparsity measures relating to an active user, a scheme based on the UMS is suggested for estimating α. Experimental results demonstrate that the proposed estimates of α, markedly, outperform the schemes for which α is kept constant across all predictions (fixed-α schemes), on accuracy of predicted ratings.     

An Approach for Deriving a Global Representation of Data Sources Having Different Formats and Structures

In this paper we propose an approach for deriving a global representation of data sources having different formats and structures. The proposed approach is based on the exploitation of a particular conceptual model for both uniformly representing such data sources and reconstructing both their intra-source and their inter-source semantics. Along with the global representation, our approach returns two support structures which improve the access transparency to stored information, namely, a set of mappings, encoding the transformations carried out during the construction of the global representation, and a set of views, allowing to obtain instances of the concepts of the global representation from instances of the concepts of the input data sources. The paper also describes a prototype which implements the proposed approach.A preliminary and partial version of this paper appears under the title A semi-automatic technique for constructing a global representation of information sources having different formats and structure in the Proceedings of the Conference Database and Expert System Applications (DEXA 2001), Munich, September 2001.     

Localized topology control for unicast and broadcast in wireless ad hoc networks

We propose a novel localized topology-control algorithm for each wireless node to locally select communication neighbors and adjust its transmission power accordingly such that all nodes together self-form a topology that is energy efficient simultaneously for both unicast and broadcast communications. We theoretically prove that the proposed topology is planar, which meets the requirement of certain localized routing methods to guarantee packet delivery; it is power-efficient for unicast - the energy needed to connect any pair of nodes is within a small constant factor of the minimum; it is also asymptotically optimum for broadcast - the energy consumption for broadcasting data on top of it is asymptotically the best among all structures constructed using only local information; it has a constant bounded logical degree, which will potentially save the cost of updating routing tables if used. We further prove that the expected average physical degree of all nodes is a small constant. To the best of our knowledge, this is the first localized topology-control strategy for all nodes to maintain a structure with all these desirable properties. Previously, only a centralized algorithm was reported. Moreover, by assuming that the node ID and its position can be represented in O(log n) bits for a wireless network of n nodes, the total number of messages by our methods is in the range of theoretical results are corroborated in the simulations.     

全局和局部信息融合的案情关键要素识别

司法人工智能中主要挑战性问题之一是案情关键要素识别,现有方法仅将案情要素作为一个命名实体识别任务,导致识别出的多数信息是无关的.另外,也缺乏对文本的全局信息和词汇局部信息的有效利用,导致要素边界识别的效果不佳.针对这些问题,提出一种融合全局和局部信息的关键案情要素识别方法.所提方法首先利用BERT模型作为司法文本的输入共享层以提取文本特征.然后,在共享层之上建立司法案情要素识别、司法文本分类(全局信息)、司法中文分词(局部信息)这3个子任务进行联合学习模型.最后,在两个公开数据集上测试所提方法的效果,结果表明:所提方法 F1值均超过了现有的先进方法,提高了要素实体分类的准确率并减少了识别边界错误问题.     

Adaptive load balancing in learning-based approaches for many-core embedded systems

Adaptive routing algorithms improve network performance by distributing traffic over the whole network. However, they require congestion information to facilitate load balancing. To provide local and global congestion information, we propose a learning method based on dual reinforcement learning approach. This information can be dynamically updated according to the changing traffic condition in the network by propagating data and learning packets. We utilize a congestion detection method which updates the learning rate according to the congestion level. This method calculates the average number of free buffer slots in each switch at specific time intervals and compares it with maximum and minimum values. Based on the comparison result, the learning rate sets to a value between 0 and 1. If a switch gets congested, the learning rate is set to a high value, meaning that the global information is more important than local. In contrast, local is more emphasized than global information in non-congested switches. Results show that the proposed approach achieves a significant performance improvement over the traditional Q-routing, DRQ-routing, DBAR and Dynamic XY algorithms.     

Survey of network coding-aware routing protocols in wireless networks

In recent times, there have been many advances in the field of information theory and wireless ad hoc network technologies. Regarding information theory progression and its connection with wireless ad hoc networks, this study presents fundamental concepts related to the application of the state-of-the-art Network Coding (NC) within wireless ad hoc networks in the context of routing. To begin with, this paper briefly describes opportunistic routing and identifies differentiation between NC-aware and NC-based routing mechanisms in wireless ad hoc networks. However, our main focus is to provide a survey of available NC-aware routing protocols that make forwarding decisions based on the information of available coding opportunities across several routes within wireless ad hoc networks. The taxonomy and characteristics of various representative NC-aware routing protocols will also be discussed. In summary, we provide a comparison of available NC-aware routing schemes and conclude that NC-aware routing techniques have several advantages over traditional routing in terms of high throughput, high reliability, and lower delay in a wireless scenario. To the best of our knowledge, this work is the first that provides comprehensive discussion about NC-aware routing protocols.     

Convex Subspace Routing (CSR): Routing via anchor-based convex virtual subspaces in sensor networks

Convex Subspace Routing (CSR) is a novel approach for routing in sensor networks using anchor-based virtual coordinates. Unlike geographical routing schemes that require physical location information of nodes, obtaining which is often difficult, error-prone and costly, the Virtual Coordinate (VC) based schemes simply characterize each node by a vector of shortest hop distances to a selected subset of nodes known as anchors. Even though VC based routing (VCR) schemes benefits from having connectivity information implicitly embedded within the VCs, VCs lack the directional information available with physical coordinates. The major issues affecting routing using VCs are addressed. Due to local minima problem in the virtual space, the VCR schemes rely on backtracking or hill climbing techniques to overcome the local minima. Convex Subspace Routing, in contrast, avoids using anchors that cause local minima. It dynamically selects subsets of anchors that define subspaces to provide convex distance functions from source to destination. Consequently, it is less sensitive to anchor placement and over anchoring, and does not require tracking route history for backtracking, resulting in shorter packet lengths and energy efficient operation. Three techniques for selection of convex subspaces are proposed and evaluated. Performance evaluation for several different network topologies indicates that CSR significantly outperforms the existing VCR scheme, Logical Coordinate Routing (LCR), while being competitive with geographic coordinate based Greedy Perimeter Stateless Routing (GPSR), even though latter makes use of node location information.     

Incremental exemplar learning schemes for classification on embedded devices

Although memory-based classifiers offer robust classification performance, their widespread usage on embedded devices is hindered due to the device’s limited memory resources. Moreover, embedded devices often operate in an environment where data exhibits evolutionary changes which entails frequent update of the in-memory training data. A viable option for dealing with the memory constraint is to use Exemplar Learning (EL) schemes that learn a small memory set (called the exemplar set) of high functional information that fits in memory. However, traditional EL schemes have several drawbacks that make them inapplicable for embedded devices; (1) they have high memory overheads and are unable to handle incremental updates to the exemplar set, (2) they cannot be customized to obtain exemplar sets of any user-defined size that fits in the memory and (3) they learn exemplar sets based on local neighborhood structures that do not offer robust classification performance. In this paper, we propose two novel EL schemes, (Entropy-Based Exemplar Learning) and (AUC-Based Exemplar Learning) that overcome the aforementioned short-comings of traditional EL algorithms. We show that our schemes efficiently incorporate new training datasets while maintaining high quality exemplar sets of any user-defined size. We present a comprehensive experimental analysis showing excellent classification-accuracy versus memory-usage tradeoffs using our proposed methods.     

Exploring probabilistic localized video representation for human action recognition

In recent years, the bag-of-words (BoW) video representations have achieved promising results in human action recognition in videos. By vector quantizing local spatial temporal (ST) features, the BoW video representation brings in simplicity and efficiency, but limitations too. First, the discretization of feature space in BoW inevitably results in ambiguity and information loss in video representation. Second, there exists no universal codebook for BoW representation. The codebook needs to be re-built when video corpus is changed. To tackle these issues, this paper explores a localized, continuous and probabilistic video representation. Specifically, the proposed representation encodes the visual and motion information of an ensemble of local ST features of a video into a distribution estimated by a generative probabilistic model. Furthermore, the probabilistic video representation naturally gives rise to an information-theoretic distance metric of videos. This makes the representation readily applicable to most discriminative classifiers, such as the nearest neighbor schemes and the kernel based classifiers. Experiments on two datasets, KTH and UCF sports, show that the proposed approach could deliver promising results.     

多QoS约束的层次多播路由算法框架

为了解决网络路由的扩展性问题。大型网络通常被划分成若干个不同的域。拓扑聚集是对这些域的拓扑状态信息进行汇总的过程。在拓扑聚集的基础上,QoS层次多播路由算法用来构造满足QoS要求的域闻多播树。现有的QoS层次多播路由算法在其拓扑聚集和路径计算的过程中都只考虑了存在两个QoS特征值的情况。本文提出了一种具有多QoS约束的层次多播路由算法框架HMRMQ（Hierarchical Multicast Routing with Multiple QoS constraints）,此算法框架不仅为基于多QoS特征值的拓扑状态聚集和状态信息表示提供了新的方法,而且提出了一种适应于多QoS约束的层次多播路由新算法。我们提出的状态信息表示法和拓扑聚集算法都具有很好的扩展性,分布式的路由算法也便于某些安全性策略的实施。理论分析和实验结果不仅证明了HMRMQ的正确性和有效性,同时也表明了HMRMQ在网络路由的扩展性、路由成功率、网络代价以及报文负载等方面都具有良好的性能。