Prediction- and simulation-error based perceptron training: Solution space analysis and a novel combined training scheme

Previous papers have noted the difficulty in obtaining neural models which are stable under simulation when trained using prediction-error-based methods. Here the differences between series-parallel and parallel identification structures for training neural models are investigated. The effect of the error surface shape on training convergence and simulation performance is analysed using a standard algorithm operating in both training modes. A combined series-parallel/parallel training scheme is proposed, aiming to provide a more effective means of obtaining accurate neural simulation models. Simulation examples show the combined scheme is advantageous in circumstances where the solution space is known or suspected to be complex. (c) 2006 Elsevier B.V. All rights reserved.

Packet classification by multi-level cutting of the classification space: An algorithmic-architectural solution for IP packet classification in next generation networks

Traditionally, the Internet provides only a “best-effort” service, treating all packets going to the same destination equally. However, providing differentiated services for different users based on their quality requirements is increasingly becoming a demanding issue. For this, routers need to have the capability to distinguish and isolate traffic belonging to different flows. This ability to determine the flow each packet belongs to is called packet classification. Technology vendors are reluctant to support algorithmic solutions for classification due to their non-deterministic performance. Although CAMs are favoured by technology vendors due to their deterministic high lookup rates, they suffer from the problems of high power dissipation and high silicon cost. This paper provides a new algorithmic-architectural solution for packet classification that mixes CAMs with algorithms based on multi-level cutting the classification space into smaller spaces. The provided solution utilizes the geometrical distribution of rules in the classification space. It provides the deterministic performance of CAMs, support for dynamic updates, and added flexibility for system designers.

Packet classification by multi-level cutting of the classification space

Traditionally, the Internet provides only a “best-effort” service, treating all packets going to the same destination equally. However, providing differentiated services for different users based on their quality requirements is increasingly becoming a demanding issue. For this, routers need to have the capability to distinguish and isolate traffic belonging to different flows. This ability to determine the flow each packet belongs to is called packet classification. Technology vendors are reluctant to support algorithmic solutions for classification due to their nondeterministic performance. Although content addressable memories (CAMs) are favoured by technology vendors due to their deterministic high-lookup rates, they suffer from the problems of high-power consumption and high-silicon cost. This paper provides a new algorithmic-architectural solution for packet classification that mixes CAMs with algorithms based on multilevel cutting of the classification space into smaller spaces. The provided solution utilizes the geometrical distribution of rules in the classification space. It provides the deterministic performance of CAMs, support for dynamic updates, and added flexibility for system designers.

The basis for the formation of stable metal clusters on an electrode surface

Metal nanoclusters can be produced cheaply and precisely in an electrochemical environment. Experimentally this method works in some systems, but not in others, and the unusual stability of the clusters has remained a mystery. We have simulated the deposition of the clusters using classical molecular dynamics and studied their stability by grand-canonical Monte Carlo simulations. We find that electrochemically stable clusters occur only in those cases where the two metals involved form stable alloys.

The genetic basis of infertility in men.

Subfertility in men is a heterogeneous syndrome, its pathophysiology remaining unknown in the majority of affected men. A large number of genes and loci are associated with sterility in experimental animals, but the human homologues of most of these genes have not been characterized. A British study suggested that, in a large proportion of men with idiopathic infertility, the disorder is inherited as an autosomal recessive trait; this provocative hypothesis needs confirmation. Because normal germ cell development requires the temporally and spatially co-ordinated expression of a number of gene products at the hypothalamic, pituitary and testicular levels, it is safe to predict that a large number of autosomal, as well as X- and Y-linked, genes will probably be implicated in different subsets of male subfertility.

Geometry in preduals of spaces of 2-homogeneous polynomials on Hilbert spaces

Let H be a (real or complex) Hilbert space. Using spectral theory and properties of the Schatten–Von Neumann operators, we prove that every symmetric tensor of unit norm in HoH is an infinite absolute convex combination of points of the form xox with x in the unit sphere of the Hilbert space. We use this to obtain explicit characterizations of the smooth points of the unit ball of HoH .

Structural basis of substrate selectivity in the glycerol-3-phosphate: phosphate antiporter GlpT.

Major facilitators represent the largest superfamily of secondary active transporter proteins and catalyze the transport of an enormous variety of small solute molecules across biological membranes. However, individual superfamily members, although they may be architecturally similar, exhibit strict specificity toward the substrates they transport. The structural basis of this specificity is poorly understood. A member of the major facilitator superfamily is the glycerol-3-phosphate (G3P) transporter (GlpT) from the Escherichia coli inner membrane. GlpT is an antiporter that transports G3P into the cell in exchange for inorganic phosphate (Pi). By combining large-scale molecular-dynamics simulations, mutagenesis, substrate-binding affinity, and transport activity assays on GlpT, we were able to identify key amino acid residues that confer substrate specificity upon this protein. Our studies suggest that only a few amino acid residues that line the transporter lumen act as specificity determinants. Whereas R45, K80, H165, and, to a lesser extent Y38, Y42, and Y76 contribute to recognition of both free Pi and the phosphate moiety of G3P, the residues N162, Y266, and Y393 function in recognition of only the glycerol moiety of G3P. It is the latter interactions that give the transporter a higher affinity to G3P over Pi.