Concurrent Cholesky factorization of positive definite banded Hermitian-matrices

First the Cholesky factorization is extended to cover uniformly partitioned banded positive definite matrices of rank n which may be real symmetric or Hermitian. Then two stratagems are given for the use of the algorithm in concurrent machines where the number of processing elements is less than required to factor the matrix in as few serial steps as possible, and where uniformly high efficiency is expected from all processing elements. Expressions are given for the efficiency factor e appearing in the speed-up expression g = eN, and these are specialized for the N node hypercube machine as a function of partition size s, the number N of processing elements of the hypercube machine, and the cost μ of interelement transmission relative to computation. It is shown that efficiency factor e is inversely proportional to μ/s, and that e is almost independent of N when N is large and μ/s = 0. The task is completed in n/s serial steps with no limit on n. The half bandwidth b of the matrix is 2Ns.     

The road to chaos by time-asymmetric Hebbian learning in recurrent neural networks

This letter aims at studying the impact of iterative Hebbian learning algorithms on the recurrent neural network's underlying dynamics. First, an iterative supervised learning algorithm is discussed. An essential improvement of this algorithm consists of indexing the attractor information items by means of external stimuli rather than by using only initial conditions, as Hopfield originally proposed. Modifying the stimuli mainly results in a change of the entire internal dynamics, leading to an enlargement of the set of attractors and potential memory bags. The impact of the learning on the network's dynamics is the following: the more information to be stored as limit cycle attractors of the neural network, the more chaos prevails as the background dynamical regime of the network. In fact, the background chaos spreads widely and adopts a very unstructured shape similar to white noise. Next, we introduce a new form of supervised learning that is more plausible from a biological point of view: the network has to learn to react to an external stimulus by cycling through a sequence that is no longer specified a priori. Based on its spontaneous dynamics, the network decides "on its own" the dynamical patterns to be associated with the stimuli. Compared with classical supervised learning, huge enhancements in storing capacity and computational cost have been observed. Moreover, this new form of supervised learning, by being more "respectful" of the network intrinsic dynamics, maintains much more structure in the obtained chaos. It is still possible to observe the traces of the learned attractors in the chaotic regime. This complex but still very informative regime is referred to as "frustrated chaos."     

Weak state versus strong state: an analysis of a probabilistic state mechanism for dynamic networks

Network protocols coordinate their decision making using information about entities in remote locations. Such information is provided by state entries. To remain valid, the state information needs to be refreshed via control messages. When it refers to a dynamic entity, the state has to be refreshed at a high rate to prevent it from becoming stale. In capacity constrained networks, this may deteriorate the overall performance of the network. The concept of weak state has been proposed as a remedy to this problem in the context of routing in mobile ad-hoc networks. Weak state is characterized by probabilistic semantics and local refreshes as opposed to strong state that is interpreted as absolute truth. A measure of the probability that the state remains valid, i.e. confidence, accompanies the state. The confidence is decayed in time to adapt to dynamism and to capture the uncertainty in the state information. This way, weak state remains valid without explicit state refresh messages. We evaluate the consistency of weak state and strong state using two notions of distortion. Pure distortion measures the average difference between the actual value of the entity and the value that is provided by the remote state. Informed distortion captures both this difference and the effect of confidence value on state consistency. Using a mathematical analysis and simulations, we show that weak state reduces the distortion values when it provides information about highly dynamic entities and/or it is utilized for protocols that is required to incur a low amount of overhead.     

DTN routing using explicit and probabilistic routing table states

Routing in communication networks involves the indirection from a persistent name (ID) to a locator. The locator specifies how packets are delivered to a destination with a particular ID. Such a mapping is provided by a routing table entry, i.e. state. In a DTN, it is hard to maintain routing state because intermittent connectivity prevents protocols from refreshing states when they become inaccurate. In prior work, per-destination state mostly corresponds to utilities, where a high utility value about a destination implies that the probability to encounter the destination for the node maintaining the state is high. This approach depends on a particular mobility pattern in which nodes that met frequently in the past are likely to encounter in the future. In this paper, we use the concept of weak state that does not rely on external messages to remain valid (Acer et al. in MobiCom ’07: proceedings of the 13th annual ACM international conference on mobile computing and networking, pp 290–301, 2007). Our weak state realization provides probabilistic yet explicit information about where the destination is located. We build Weak State Routing protocol for Delay Tolerant Networks (WSR-D) that exploits the direction of node mobility in forwarding. It provides an osmosis mechanism to disseminate the state information to the network. With osmosis, a node has consistent information about a portion of the nodes that are located in regions relevant to its direction of mobility. Through simulations, we show that WSR-D achieves a higher delivery ratio with smaller average delay, and reduces the number of message transfers in comparison to Spray & Wait (Spyropoulos et al. in Proceedings of ACM SIGCOMM 2005 workshops: conference on computer communications, pp 252–259, 2005) and Spray & Focus (Spyropoulos et al. in IEEE/ACM Trans Netw, 16(1):77–90, 2008), a stateless and a utility based protocol, respectively.     

Effects of material non-linearity on the residual stresses in a dendritic silicon crystal ribbon

Thermal stresses developed in a dendritic silicon crystal ribbon have been shown to cause plastic deformation and residual stresses in the ribbon. This paper presents an implementation of a numerical model proposed for thermo-elasto-plastic behaviour of a material. The model has been used to study the effects of plasticity of silicon on the residual stresses. The material properties required to implement this model are all assumed, and the response of the material to the variations in these assumed parameters of the constitutive law and in the finite element mesh is investigated. The steady state growth process is observed to be periodic with non-zero residual stresses. Numerical difficulties are also encountered in the computer solution process, resulting in sharp jumps and large oscillations in the stress responses.     

Prevention of acute allograft rejection by antibody targeting of TIRC7, a novel T cell membrane protein

The in vitro activity of trovafloxacin, a new fluoroquinolone, was compared with that of ciprofloxacin, ofloxacin, fleroxacin, ceftazidime, piperacillin/tazobactam, and meropenem against 613 consecutively recovered blood isolates from recently hospitalized patients. Susceptibility testing was performed by agar dilution according to NCCLS guidelines. Test strains included Acinetobacter species (n = 26), Escherichia coli (n = 137), Enterobacter species (n = 27), Klebsiella species (n = 42), Proteus species (n = 16), Pseudomonas aeruginosa (n = 28), Serratia marcescens (n = 13), Stenotrophomonas maltophilia (n = 7), enterococci (n = 54), coagulase-negative staphylococci (n = 38), Staphylococcus aureus (n = 137), Streptococcus pneumoniae (n = 27), beta-haemolytic streptococci (n = 13), and viridans group streptococci (n = 48). The overall respective MICs at which 50% and 90% of isolates were inhibited (MIC50s and MIC90s) were as follows: trovafloxacin, 0.06 and 1 mg/l; ciprofloxacin, 0.25 and 4 mg/l; ofloxacin, 0.5 and 4 mg/l; fleroxacin, 0.5 and 16 mg/l; ceftazidime, 2 and 128 mg/l; piperacillin/tazobactam, 2 and 8 mg/l; meropenem, 0.06 and 4 mg/l. For the quinolones, the rank order of activity against gram-negative microorganisms was ciprofloxacin > trovafloxacin > ofloxacin = fleroxacin, against gram-positive organisms, trovafloxacin > ciprofloxacin = ofloxacin > fleroxacin. Data obtained showed the similar activity of trovafloxacin and ciprofloxacin against gram-negative pathogens and the superior activity of trovafloxacin against gram-positive bacteria thus making it a potential candidate for the empiric treatment of patients with suspected bacteremia and sepsis.     

On nonlinear finite element analysis in single-, multi- and parallel-processors

Numerical solution of nonlinear equilibrium problems of structures by means of Newton-Raphson type iterations is reviewed. Each step of the iteration is shown to correspond to the solution of a linear problem, therefore the feasibility of the finite element method for nonlinear analysis is established. Organization and flow of data for various types of digital computers, such as single-processor/single-level memory, single-processor/two-level-memory, vector-processor/two-level-memory, and parallel-processors, with and without sub-structuring (i.e. partitioning) are given. The effect of the relative costs of computation, memory and data transfer on substructuring is shown. The idea of assigning comparable size substructures to parallel processors is exploited. Under Cholesky type factorization schemes, the efficiency of parallel processing is shown to decrease due to the occasional shared data, just as that due to the shared facilities.     

Effects of compression on water distribution in gas diffusion layer materials of PEMFC in a point injection device by means of synchrotron X-ray imaging

In this study, ex-situ experiments performed with a point injection device are conducted to evaluate water distributions in gas diffusion layer (GDL) materials which serve as porous transport media in polymer electrolyte membrane fuel cells (PEMFCs). In this regard, GDL samples manufactured by SGL Group are placed into the point injection device and visualized by means of synchrotron X-ray radiographic and tomographic imaging. The resulting image data undergoes a coordinate transformation that ascertains water agglomerations in GDL pores with regard to their radial displacements from the injection point. In this way, water transport in two different GDL samples possessing the same structural characteristics, but with unique compression rates, are investigated in terms of in-plane water distribution. The radial displacement analysis indicated that the pore saturation of the compressed GDL is higher in both the micro porous layer (MPL) region and the carbon fiber substrate region than that of the uncompressed GDL. The water agglomerations in the uncompressed GDL are predominantly observed in the vicinity of the injection point, indicating a limited in-plane transport. Conversely, in the compressed case water accumulations are detected far from the injection point, even at the edge of the GDL, pointing out that compression promotes the in-plane transport. Prior to the ex-situ experiments, both GDL samples have undergone an ageing procedure to mimic realistic cell operating conditions.     

Neural network-supported patient-adaptive fall prevention system

Neural Computing and Applications - Patient falls due to unattended bed-exits are costly to patients, healthcare personnel and hospitals. Numerous researches based on up to three predetermined...