Search

Learning human mobility behaviors from location-sensing data are crucial to mobility data mining because of its potential to address a range of analytical purposes in mobile context reasoning, including exploration, inference, and prediction. However, existing approaches suffer from two practical problems: temporal and spatial sparsity. To address these shortcomings, we present two unsupervised learning methods to model the mobility behaviors of multiple users (i.e., a population), considering efficiency and accuracy. These methods intelligently overcome the sparsity in individual data by seeking temporal commonality among users' heterogeneous location behaviors. The advantages of our models are highlighted through experiments on several real-world mobility data sets, which also show how our methods can realize the three analytical purposes in a unified manner. [ABSTRACT FROM AUTHOR]

The Pseudo Trust (PT) protocol is designed, in which each peer generates an unforgeable and verifiable pseudonym by using a one-way hash function. The results of trace-driven simulations have shown that PT is scalable in both static and dynamic environments.

A novel incentive-based scheduling scheme for grid computing is presented. Results suggest the achievement of highly successful job execution and fair profit allocation.

An interoverlay optimization scheme (IOO) is proposed for providing scalable live streaming services. Findings reveal the excellent performance of IOO in terms of resource utilization and the quality of service (QoS) of streaming services.

A progressive and probabilistic approach is presented for the effective protection of sensitive information and for preserving privacy for users and service providers. Findings reveal the efficient running of the approach on PDAs.

A distributed IP network emulator cluster EMPOWER ( Emulation of the performance of wide area networks ) is presented. It can be used to emulate a large network with a limited number of commodity computers and can generate user-defined arbitrary network conditions and traffic dynamics at packet level for specific test scenarios.

A discussion of cut-through switching for high speed local area networks is presented, focusing on adaptive-trail routing. Topics include the up/down routing algorithm, the smart-routing algorithm, routing tables, adaptivity, network workload, and network latency.

The multicast tree implementation was studied to support the wormhole routed multistage networks. An optimal software multicast algorithm was used for multistage interconnection networks supporting one-port communication. Minimum multicast latency is always achieved regardless of the message size or startup latency proving its superiority over other application-level broadcast services.

Switch-based or indirect networks have potential in network architecture for constructing scalable parallel computers (SPCs). The majority are based on multistage interconnection networks (MINs), and a study compares four different wormhole MINs: traditional MINs, dilated MINs, bidirectional MINs and virtual channel MINs. It concludes that a two-dilated MIN can outperform a corresponding bidirectional MIN for the majority of traffic conditions, and is more suitable for scalable parallel computers.

A method, based on the turn model for creating wormhole routing algorithms, facilitates the designing of a fault-tolerant routing model that can handle dynamic faults. One-fault-tolerant and (n - 1)-fault-tolerant routing algorithms for wormhole-routed meshes without virtual channels have been designed. Previous methods for devising such algorithms have been reviewed. Merits and demerits of fault-routing algorithms are discussed.

Memory access degradation (MAD) kernels efficiently determine the shared memory performance of different types of memory access workloads. The performance measurements are independent of other algorithms or applications. Measurement of the performance of Sequence Symmetry and BBN TC2000 multiprocessors indicate that the static characterization parameter for the TC2000 is higher than the Symmetry. MAD kernels can be used independently to determine the sensitivity of shared memory organization or in conjunction with SAD kernels to evaluate memory overhead disturbances.

Machine learning has been highly successful in data-intensive applications but is often hampered when the data set is small. Recently, Few-shot Learning (FSL) is proposed to tackle this problem. Using prior knowledge, FSL can rapidly generalize to new tasks containing only a few samples with supervised information. In this article, we conduct a thorough survey to fully understand FSL. Starting from a formal definition of FSL, we distinguish FSL from several relevant machine learning problems. We then point out that the core issue in FSL is that the empirical risk minimizer is unreliable. Based on how prior knowledge can be used to handle this core issue, we categorize FSL methods from three perspectives: (i) data, which uses prior knowledge to augment the supervised experience; (ii) model, which uses prior knowledge to reduce the size of the hypothesis space; and (iii) algorithm, which uses prior knowledge to alter the search for the best hypothesis in the given hypothesis space. With this taxonomy, we review and discuss the pros and cons of each category. Promising directions, in the aspects of the FSL problem setups, techniques, applications, and theories, are also proposed to provide insights for future research. [ABSTRACT FROM AUTHOR]

An architecture-independent processor mapping method to minimize the amount of data transferred between processor memories for 'Block' to 'Cyclic(c)' or vice-versa redistributions of arbitrary number dimensions is introduced. The procedure, which maps logical processors to data elements, provides a 40% improvement to the data redistribution time for data set sizes over other traditional techniques. This was revealed when it was run on an IBM SP processor at Argonne National Laboratory.

A novel technique developed to observe routine devices in wormhole-routed networks aims to create a conducive condition for deadlock-free routing which at the same time also would be deadlock-immune. This message flow model can be used to show the deadlock free character of a wormhole routing mechanism for which the CAD based models can be applied. This model helps the precise movement of message, so it can be suitable to a broad range of routing device. In this system messages are categorized into different subsets and channelized. It also helps the 2-D mesh by arranging a nonminimal-path adaptive multicast routing device in 2-D meshes.

A wormhole routed n-dimensional mesh is used for the efficient implementation of multicast communication services for scalable multiprocessors. The mesh includes several commercial machines, and the U-cube and U-mesh multicast algorithm yield contention-free minimum-time multicast trees. The wormhole routed method is implemented on a 64-node nCUBE-2 and a 168-node Symult 2010. This method is more efficient than other conventional methods.

Two-dimensional mesh multicomputers use deadlock-free multicast algorithms that employ wormhole routing. These algorithms lose their deadlock-free characteristics when they contain multicast traffic. The algorithms multipath routing reveals the presence of hot spots detrimental to communication performance. Hyperboles and general n-dimensional meshes also use multicast algorithms.

Data dependence uniformization is a loop parallelization technique used to eliminate difficulty involved in parallelizing nested loops having complex dependence constraints. A simple set of basic dependences which comprises all dependence constraints is formulated. It can be appended to every iteration to substitute all original dependences, making it uniform. Index synchronization is proposed to follow the basic dependence vectors in the uniformized dependence patterns.