Alpha matting is an important topic in areas of computer vision. It has various applications, such as virtual reality, digital image and video editing, and image synthesis. The conventional approaches for alpha matting perform unsatisfactorily when they encounter complicated background and foreground. It is also difficult for them to extract alpha matte accurately when the foreground objects are transparent, semi-transparent, perforated or hairy. Fortunately, the rapid development of deep learning techniques brings new possibilities for solving alpha matting problems. In this thesis, we propose a residual convolutional grid network for alpha matting, which is based on the convolutional neural networks (CNNs) and can learn the alpha matte directly from the original image and its trimap. Our grid network consists of horizontal residual convolutional computation blocks and vertical upsampling/downsampling convolutional computation blocks. By choosing different paths to pass information by itself, our network can not only retain the rich details of the image but also extract high-level abstract semantic information of the image. The experimental results demonstrate that our method can solve the matting problems that plague conventional matting methods for decades and outperform all the other state-of-the-art matting methods in quality and visual evaluation. The only matting method performs a little better than ours is the current best matting method. However, that matting method requires three times amount of trainable parameters compared with ours. Hence, our matting method is the best considering the computation complexity, memory usage, and matting performance.

Modern cellular networks including LTE (Long Term Evolution) and the evolving LTE- Advanced provide high-speed and high-capacity data services for mobile users. As we become more reliant on wireless connectivity, the security of voice and data transmissions on the network becomes increasingly important. While the LTE network standards provide strict security guidelines, these requirements may not be completely followed when LTE networks are deployed in practice. This project provides a method for improving the security of LTE networks by 1) characterizing a gap between security requirements defined in the standards and practical implementations, 2) designing a language to express the encoding formats of one of LTE’s network-layer protocols, 3) developing a compiler to translate a protocol description in our language into an implementation, and 4) providing recommendations on lessons learned during development of the language and compiler to support development of future protocols that employ formal representations. In this way, our work demonstrates how a formal language can be utilized to represent a cellular network protocol and serves as an example for further research on how adding formalism to network standards can help ensure that the security goals defined in the standards can be upheld in an implementation.

Zhang Li. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2004. / Includes bibliographical references (leaves 71-74). / Abstracts in English and Chinese. / Abstract --- p.ii / 摘要 --- p.iv / Acknowledgments --- p.v / Contents --- p.vi / Chapter Chapter 1 --- Introduction --- p.1 / Chapter Chapter 2 --- Internet Routing and Performance Studies --- p.5 / Chapter 2.1 --- Border Gateway Protocol Version 4 (BGP4) --- p.5 / Chapter 2.2 --- Internet Performance Studies --- p.6 / Chapter 2.3 --- Improve Routing Performance --- p.8 / Chapter 2.3.1 --- Traffic Engineering --- p.8 / Chapter 2.3.2 --- Network-layer Techniques --- p.9 / Chapter 2.3.3 --- Minimum Delay Routing --- p.11 / Chapter Chapter 3 --- Overlay-based Techniques --- p.15 / Chapter 3.1 --- Content Distribution Network --- p.16 / Chapter 3.2 --- Relative Overlay Techniques --- p.16 / Chapter Chapter 4 --- Overlay Minimum Delay Routing --- p.19 / Chapter 4.1 --- Minimum Delay Routing in Overlay Network --- p.21 / Chapter 4.1.1 --- Problem Formulation --- p.21 / Chapter 4.1.2 --- Necessary and Sufficient Conditions for Distributed Computing --- p.25 / Chapter 4.1.3 --- Optimal Overlay Auxiliary Routing (00AR) Algorithm --- p.29 / Chapter 4.2 --- Performance Comparing with Optimal IP routing --- p.31 / Chapter Chapter 5 --- Sub-Optimal Overlay Auxiliary Routing Algorithm --- p.39 / Chapter 5.1 --- Approximation Conditions to Optimal Overlay Routing --- p.40 / Chapter 5.2 --- SOAR algorithm Overview --- p.44 / Chapter 5.3 --- Distributing Traffic over Multiple Paths --- p.49 / Chapter 5.3.1 --- Adaptive load Adjustment --- p.50 / Chapter 5.3.2 --- Per-flow Routing Control --- p.53 / Chapter 5.4 --- Discussion on Marginal Delay of an OAR Link --- p.55 / Chapter Chapter 6 --- Performance Evaluation of Sub-Optimal Overlay Auxiliary Routing --- p.57 / Chapter 6.1 --- Experiment Method Description --- p.57 / Chapter 6.2 --- Comparison of overall delay --- p.61 / Chapter 6.3 --- Effect of the Routing update interval parameters xx and yy --- p.64 / Chapter 6.4 --- Comparison of packet loss rate --- p.66 / Chapter 6.5 --- Comparison of potential maximum data transmission rate --- p.68 / Chapter 6.6 --- Stability of the OAR load-balancing heuristics algorithm --- p.69 / References --- p.71 / Appendix --- p.75

Graph theory is the foundation on which social network analysis (SNA) is built. With the flood of "big data," graph theoretical concepts and their linear algebraic counterparts are essential tools for analysis in the burgeoning field of network data analysis, in which SNA is a subset. Here we begin with an overview of SNA. We then discuss the common descriptive measures taken on network data as well as proposing new measures specific to homeless networks. We also define a new data structure which we call the location sequence matrix. This data structure makes certain computational network analyses particularly easy. Finally we apply Pulse Processes in a new way to the homeless network in Utah. We believe the new data structure and pulse processes, when used for analysis of the Utah homeless services. In particular, pulse processes, first introduced by Brown, Roberts, and Spencer, to analyze energy demand, form a dynamic population model that can provide a measure of the stability in a network and the patterns of action of individuals experiencing homelessness.

Thesis (M.S.)--University of Missouri--Rolla, 2007. / Vita. The entire thesis text is included in file. Title from title screen of thesis/dissertation PDF file (viewed December 6, 2007) Includes bibliographical references (p. 60-63).

<p><p>Title: Network in the knowledge transfer process</p><p>Authors: Ying-Sze Tang (Virginia), Pattamawan Chaikiturajai (Koy)</p><p>Supervisor: Sven-Olof Yrjö Collin</p><p>Course: Dissertation 15 ECTS, spring 2010</p><p>Seminar date: June 1, 2010</p><p>Key words: Knowledge transfer, network, knowledge transfer by network</p><p>Purpose: To examine the roles of network in the knowledge transfer process</p><p>Method: The aim is to examine the roles of network in the knowledge transfer process by using theories as a foundation and then relating literatures with empirical data. An abductive approach and qualitative case study approach would be conducted.</p><p>Theory: The theoretical framework of reference consists of knowledge transfer, network and knowledge transfer by network. The conceptual models which are developed by the authors will be described at the end of chapter 3.Empirical method: Secondary data such as companies‘ information from website and annual report are used for data collection. Since the comparison could be made, this study is to perform interviews with two international companies which are HMS Industrial Networks AB and Synovate Limited. The interviewees are the persons who participated in the communication with overseas people.Analysis: The analysis part consists of two sections. The first section is a within-case analysis of HMS Industrial Networks AB and Synovate Limited while the second section is a cross-case analysis of both companies.</p><p>Conclusion: The major finding after doing this thesis is that all main networks including home country network, host country network and subsidiaries network are crucial factors in knowledge transfer process. Home country network acts as an ‗Advisor‘ to get assistance and guidance for gaining knowledge. Host country network acts as a ‗Punching Tool‘ which helps company to enter into an unfamiliar and difficult foreign market due to language and culture <em>barrier. Subsidiaries network acts as a ‗Knowledge Booster‘ which help each other about transferring knowledge between host countries to enhance the whole process. It is because business operations are conducted in subsidiaries locally but not in the headquarters.</em></p></p>

We model Pup Matching, the logistics problem of matching or pairing semitrailers known as pups to cabs able to tow one or two pups simultaneously, as an NP-complete version of the Network Loading Problem (NLP). We examine a branch and bound solution approach tailored to the NLP formulation through the use of three families of cutting planes and four heuristic procedures. Theoretically, we specify facet defining conditions for a cut family that we refer to as odd flow inequalities and show that each heuristic yields a 2-approximation. Computationally, the cheapest of the four heuristic values achieved an average error of 1.3% among solved test problems randomly generated from realistic data. The branch and bound method solved to optimality 67% of these problems. Application of the cutting plane families reduced the average relative difference between upper and lower bounds prior to branching from 18.8% to 6.4%. / Singapore-MIT Alliance (SMA)

Conguring a network system to operate at optimal levels of performance re-quires a comprehensive understanding of the eects of a variety of system parameterson crucial metrics like connectivity and resilience to network attacks. Traditionally,omni-directional antennas have been used for communication in wireless sensor net-works. In this thesis, a hybrid communication model is presented where-in, nodes ina network are capable of both omni-directional and uni-directional communication.The eect of such a model on performance in randomly deployed wireless sensor net-works is studied, specically looking at the eect of a variety of network parameterson network performance.The work in this thesis demonstrates that, when the hybrid communication modelis employed, the probability of 100% connectivity improves by almost 90% and thatof k-connectivity improves by almost 80% even at low node densities when comparedto the traditional omni-directional model. In terms of network security, it was foundthat the hybrid approach improves network resilience to the collision attack by almost85% and the cost of launching a successful network partition attack was increased byas high as 600%. The gains in connectivity and resilience were found to improve withincreasing node densities and decreasing antenna beamwidths.

Title: Network in the knowledge transfer process Authors: Ying-Sze Tang (Virginia), Pattamawan Chaikiturajai (Koy) Supervisor: Sven-Olof Yrjö Collin Course: Dissertation 15 ECTS, spring 2010 Seminar date: June 1, 2010 Key words: Knowledge transfer, network, knowledge transfer by network Purpose: To examine the roles of network in the knowledge transfer process Method: The aim is to examine the roles of network in the knowledge transfer process by using theories as a foundation and then relating literatures with empirical data. An abductive approach and qualitative case study approach would be conducted. Theory: The theoretical framework of reference consists of knowledge transfer, network and knowledge transfer by network. The conceptual models which are developed by the authors will be described at the end of chapter 3.Empirical method: Secondary data such as companies‘ information from website and annual report are used for data collection. Since the comparison could be made, this study is to perform interviews with two international companies which are HMS Industrial Networks AB and Synovate Limited. The interviewees are the persons who participated in the communication with overseas people.Analysis: The analysis part consists of two sections. The first section is a within-case analysis of HMS Industrial Networks AB and Synovate Limited while the second section is a cross-case analysis of both companies. Conclusion: The major finding after doing this thesis is that all main networks including home country network, host country network and subsidiaries network are crucial factors in knowledge transfer process. Home country network acts as an ‗Advisor‘ to get assistance and guidance for gaining knowledge. Host country network acts as a ‗Punching Tool‘ which helps company to enter into an unfamiliar and difficult foreign market due to language and culture barrier. Subsidiaries network acts as a ‗Knowledge Booster‘ which help each other about transferring knowledge between host countries to enhance the whole process. It is because business operations are conducted in subsidiaries locally but not in the headquarters.

Embedding of virtual network (VN) requests on top of a shared physical network poses an intriguing combination of theoretical and practical challenges. Two major problems with the state-of-the-art VN embedding algorithms are their indifference to the underlying substrate topology and their lack of re-optimization mechanisms for already embedded VN requests. We argue that topology-aware embedding together with re-optimization mechanisms can improve the performance of the previous VN embedding algorithms in terms of acceptance ratio and load balancing. The major contributions of this thesis are twofold: (1) we present a mechanism to differentiate among resources based on their importance in the substrate topology, and (2) we propose a set of algorithms for re-optimizing and re-embedding initially-rejected VN requests after fixing their bottleneck requirements. Through extensive simulations, we show that not only our techniques improve the acceptance ratio, but they also provide the added benefit of balancing load better than previous proposals. The metrics we use to validate our techniques are improvement in acceptance ratio, revenue-cost ratio, incurred cost, and distribution of utilization.