Lightweight probabilistic broadcast

Each device in being mobile ad hoc network (MANET) is overcome to navigate independently in any direction of side. The key challenge in building a MANET is availing every device to frequently maintain the information using static and dynamic topology. When the mobility is high scalable broadcast algorithm (SBA) is used and when the mobility is low memory aided broadcast algorithm (MaBA) is used for broadcasting packets. The MaBA is the algorithm where hubs on their memory to choose whether to forward stream or backward stream. The fuzzy enabled environs aware broadcast algorithm; a host maintains the local mobility around itself and defined user to choose which of two broadcast mechanisms to use. The Fuzzy logic is a Boolean conventional logic which has been increased to handle the concept of partial values between "completely true" and "completely false". As its name convinced different logic underlying the modes of reasoning which are approximate rather than exact. The significance of fuzzy logic from the host derives. The fuzzy logic is going to work with the two algorithms SBA and MaBA and it divides the network into static, semi-static and dynamic.

In Delay Tolerant Networks (DTNs), message forwarding is a challenging open problem because of low node density and lack of global information. Most of the previous forwarding methods select a node as a relay node with the best cumulative and global ability of contacting the destination without considering the maximum message acceptable delivery latency. However, if the network is composed of several communities, the cumulatively and globally calculated forwarding metric (and not in each community) will not have suitable estimation of contact between relay node and destination. Therefore, in this paper, we introduce and evaluate a novel forwarding method, which called Community-based Contact Prediction message forwarding method (CCPMFM). The CCPMFM exploits a semi-Markov process to explain and predict the node mobility between communities. The CCPMFM selects the relay node with regard to the maximum message acceptable delivery latency. To calculate the forwarding metric, CCPMFM considers the contact patterns of each community only for the same community and not for the entire network. CCPMFM also selects the relay node with regard to maximum message acceptable delivery latency. The evaluation results show that this method improves the delivery ratio and reduces the number of message forwardings and delivery latency with respect to prevalent DTNs forwarding methods.

In challenged networks, such as opportunistic networks or Delay Tolerant Networks (DTNs), due to frequent network partition an end-to-end connection between source and destination may not always be possible due to nodes move unpredictably with high speeds. For such type of networks, traditional routing protocols fails to transmit and also no guarantee on delivery of data to specified destination because these protocols deliver data definitely whenever there exists continuous end-to-end connectivity between source and destination over time. To provide efficient routing for such networks replication based routing schemes were proposed to achieve higher delivery probability. Unlike Internet protocols, Delay Tolerant Networks routing protocols follow store-carryforward mechanism to reduce packets dropping probability, tolerate delays and improve delivery performance in terms of delivery ratio and delivery latency. This paper defines the most efficient replication based routing protocols such as epidemic, PROPHET, spray and wait and Encounter based routing. This paper also explains how message TTL, buffer sizes and area of grid impact performance of replication based routing protocol in terms of delivery ratio and Average latency and overhead ratio.

Broadcasting in Ad hoc networks is an important information dissemination way. However, flooding may give rise to broadcast storm problem which causes severe contention, collision and delay. This paper proposes a dynamic probabilistic broadcast algorithm based on neighbour discovery that ensures the higher reliability at lower retransmission. The dynamic probabilistic broadcast algorithm forwards the information from the node (INode) that has obtained it to the node (UNode) that has not received it. By counting the number of UNode and INode of neigbors, INodes dynamically adjust the probability of broadcasting to reduce the number of information replication but cover more UNodes. Dynamic probabilistic algorithm can not only effectively reduce the number of retransmission, but also ensure the effectiveness and reliability. Simulation results show that the number of broadcasting of this algorithm is significantly lower than other algorithms.

An integrated software environment for large-scale Occam programming

Disruption Tolerant Networks (DTNs) are characterized by frequent network partitioning. This causes network connectivity to be opportunistic. In recent years, many collaborative applications have been envisioned for DTNs. These applications rely on the availability of reliable group communication protocols. To address this challenge in DTNs, we propose a scalable multicasting scheme that deterministically guarantees message delivery to all receivers in the group. Also, our scheme makes controlled use of non-multicast nodes to reduce latency of message delivery. Our work introduces a new measure called termination delay, which is the delay incurred in assuring that the receivers have obtained a copy of the message. Simulation results show that the proposed scheme has much smaller delivery latency and termination delay than other schemes in literature.

In wireless multi-hop networks, broadcasting is required for many purposes to disseminate a message to all hosts in the network, such as routing and data dissemination. Flooding is the simplest form of broadcasting, however it causes the 'broadcast storm' problem in which redundant broadcast and scheduling are two core subproblems to be solved. In this paper, we address the problem of redundant broadcast and design a new completely localized algorithm, i.e, the decision is made based on the local topology information only. The advantage of this algorithms is that it can adapt to the topology change quickly than other non-localized algorithms due to its locality while achieve even better performance in efficiency than the best known non-localized algorithm does.

Limited energy is a big challenge for large scale wireless sensor networks (WSN). Previous research works show that modulation scaling is an efficient technique to reduce energy consumption. However, the impacts of using modulation scaling on packet delivery latency and loss are not considered, which may have adverse effects on the application qualities. In this paper, we study this problem and propose control schemes to minimize energy consumption while ensuring application qualities. We first analyze the relationships of modulation scaling and energy consumption, end-to-end delivery latency and packet loss ratio. With the analytical model, we develop a centralized control scheme to adaptively adjust the modulation levels, in order to minimize energy consumption and ensure the application qualities. To improve the scalability of the centralized control scheme, we also propose a distributed control scheme. In this scheme, the sink will send the differences between the required and measured application qualities to the sensors. The sensors will update their modulation levels with the local information and feedback from the sink. Experimental results show the effectiveness of energy saving and QoS guarantee of the control schemes. The control schemes can adapt efficiently to the time-varying requirements on application qualities.

The paper presents experimental data obtained from measurements of interprocess group communication under PVM environment. Two main parameters: message throughput and message delivery latency are evaluated and their potential influence on execution time of applications requiring frequent distribution of large amounts of data is discussed.

Several real-time applications (e.g., video conferencing, remote control systems) demand guarantees on the message delivery latency. Supporting these QoS connections requires the existence of a routing mechanism, which computes paths that satisfy QoS delay constraints (e.g., end-to-end delays or failure recovery delays). In mobile ad hoc network, wireless links tend to frequently fail as nodes move in and out of transmission range of one another. Providing fault tolerance with QoS guarantees in such a network is challenging. The proposed new algorithm, also called a segmented backup routing (SBR), constructs a set of segmented backup paths. Each segmented backup path, protects a segment of the primary path rather than the entire path. The most important points that are made in this algorithm are two. First, one is able to identify backup paths for any selected primary path, as long as there exists a pair of node disjoint paths from source to destination. In other words, if there are two node disjoint paths, one is not forced to use either of them to be the primary path. This has the advantage that the primary path may be selected based on QoS considerations rather than a consideration of fault tolerance. The second advantage is that there are multiple backup paths, each of which protects breakdown of one or more nodes on the primary path. A backup path may be selected based on QoS considerations. In this paper, we have primarily focused on end-to-end delay as the QoS parameter. We apply our scheme to the AODV protocol.

Wireless sensor networks (WSNs) have significant limitations in available bandwidth and energy. The limited bandwidth in sensor networks can cause higher message delivery latency. In this demo, we demonstrate working of novel Zorder based data compression schemes (Z- compression) to reduce energy and save bandwidth without increasing the message delivery latency using a TelosB motes based sensor network testbed. Instead of using the popular Huffman tree style based encoding, Z-compression uses Z-order encoding to map the multidimensional sensing data into one-dimensional binary stream transmitted using a single packet. We have also designed and developed a data concatenating algorithm which can concatenate small packets into large packets, thus increases the throughput of the WSNs.

\n The rising popularity of generative models together with the growing need for flexible and exact inferences have motivated the machine learning community to look for expressive yet tractable probabilistic models. Probabilistic circuits (PCs) are a family of tractable probabilistic models capable of answering a wide range of queries exactly and in polynomial time. Their operational syntax in the form of a computational graph and their principled probabilistic semantics allow their parameters to be estimated by the highly scalable and efficient optimization techniques used in deep learning. Importantly, tractability is tightly linked to constraints on their underlying graph: by enforcing certain structural assumptions, queries like marginals, maximum a posteriori or entropy become linear time computable while still retaining great expressivity. While inference is usually straightforward, learning PCs that both obey the needed structural restrictions and exploit their expressive power has proven a challenge. Current state-of-the-art structure learning algorithms for PCs can be roughly divided into three main categories. Most learning algorithms seek to generate a usually tree-shaped circuit from recursive decompositions on data, often through clustering and costly statistical (in)dependence tests, which can become prohibitive in higher dimensional data. Alternatively, other approaches involve constructing an intricate network by growing an initial circuit through structural preserving iterative methods. Besides depending on a sufficiently expressive initial structure, these can possibly take several minutes per iteration and many iterations until visible improvement. Lastly, other approaches involve randomly generating a probabilistic circuit by some criterion. Although usually less performant compared to other methods, random PCs are orders of magnitude more time efficient. With this in mind, this dissertation aims to propose fast and scalable random structure learning algorithms for PCs from two different standpoints: from a logical point of view, we efficiently construct a highly structured binary PC that takes certain knowledge in the form of logical constraints and scalably translate them into a probabilistic circuit; from the viewpoint of data guided structure search, we propose hierarchically building PCs from random hyperplanes. We empirically show that either approach is competitive against state-of-the-art methods of the same class, and that their performance can be further boosted by simple ensemble strategies.\n

Three methods with varying degrees of flexibility are investigated: (1) Mail systems with syntax-directed naming which identify users by names that are syntactically structured according to user locations. Algorithms for load balancing among mail servers, system reconfiguration, and efficient message delivery are developed and tested using simulation. (2) Mail systems with limited location-independent access which allow users to access them from one primary location and a number of secondary locations. Procedures that keep track of users who migrate from their primary location and redirect their mail are presented. (3) The attribute-based mail system which provides maximum flexibility to users by allowing them to identify one or more mail recipients by attributes instead of only by precise names. It can also be used in mass distribution of electronic mail. An algorithm for efficient broadcasting and searching using a minimum-weight spanning tree is investigated. Criteria for evaluating electronic mail systems are presented. >

도로포장의 파손과정에는 다양하고 복합적인 원인에서 비롯되는 불확실성이 포함되어 있어 정확한 해석이 쉽지 않다. 이로 인해 최근에는 결정론적 모형보다는 확률이론이 보다 많이 활용되고 있으나, 파손의 전체적 특성만을 설명하는 일반적인 분석방안으로는 포장파손특성의 변화과정에 대해 구체적인 정보를 제공하기 어렵다. 이에 본 연구에서는 포장파손과정을 상태와 시간기준으로 분해함으로써 지역적으로 이질성을 띄는 포장파손속도와 그 분산에 대해 구체적으로 파악하고자 하였다. 또한, 분해된 확률과정을 다시 조합하는 과정을 통해 포장의 기대수명과 불확실성을 예측해 보았다. 실증분석을 위해 일반국도포장관리시스템에서 2003년부터 2010년까지 수집된 균열률 자료를 활용하였다. 이러한 시도들은 자산관리의 주요기법 중 하나인 생애주기비용분석의 신뢰성을 높일 수 있으며, 파손속도의 변화과정에 대한 이해가 필수적인 예방적 유지보수전략에 관한 기반연구로써도 중요한 의의가 있다. Precise analysis on deterioration processes of road pavements is not so simple matter due to severe uncertainty originated from a lot of explanatory variables engaged in. For those reasons, most analytical models for pavement deterioration prediction have often preferred to probabilistic approaches than deterministic models. However, the general probabilistic approaches that treat overall characteristics of population or entire sample would not be suitable for providing detail or localized information on their changing process. Considering the aspects, this paper aimed to suggest a stochastic disaggregation method to analyze the localized deterioration speeds and its variances changed by time and condition states. In addition, life expectancies and their uncertainty were estimated by probabilistic algorithm using the disaggregated stochastic process. For an empirical study, pavement inspection data (crack) accumulated from 2003 to 2010 from Korean national highway network was applied. This study can contribute to securing reliability of life cycle cost analysis, which is one of the primary analyses in road asset management, with much advanced deterioration forecasting functions. In addition, it would be meaningful trials as fundamental research for preventive maintenance strategy that demands essential understanding on changing process of the deterioration speed of pavement.

In this chapter, we study the first application of dissimilarity metrics in wireless multi-hop networks. The idea is to apply the introduced dissimilarity metrics to well-known probabilistic broadcasting algorithms based on the Euclidean distance.