Glassy behavior in neural network models of associative memory

Neural network models of associative memory exhibit a large number of spurious attractors of the network dynamics which are not correlated with any memory state. These spurious attractors, analogous to "glassy" local minima of the energy or free energy of a system of particles, degrade the performance of the network by trapping trajectories starting from states that are not close to one of the memory states. Different methods for reducing the adverse effects of spurious attractors are examined with emphasis on the role of synaptic asymmetry. (C) 2002 Elsevier Science B.V. All rights reserved.

Quantum spin models with exact dimer ground states

Inspired by the exact solution of the Majumdar-Ghosh model, a family of one-dimensional, translationally invariant spin Hamiltonians is constructed. The exchange coupling in these models is antiferromagnetic, and decreases linearly with the separation between the spins. The coupling becomes identically zero beyond a certain distance. It is rigorously proved that the dimer configuration is an exact, superstable ground-state configuration of all the members of the family on a periodic chain. The ground state is twofold degenerate, and there exists an energy gap above the ground state. The Majumdar-Ghosh Hamiltonian with a twofold degenerate dimer ground state is just the first member of the family. The scheme of construction is generalized to two and three dimensions, and illustrated with the help of some concrete examples. The first member in two dimensions is the Shastry-Sutherland model. Many of these models have exponentially degenerate, exact dimer ground states.

Study of the dynamics of protein folding through minimalistic models

Starting with the Levinthal paradox, a brief introduction to the protein folding problem is presented. The existing theories of protein folding, including the folding funnel scenario, are discussed. After briefly discussing different simulation studies of model proteins, we discuss our recent work on the dynamics of folding of the model HP-36 (the chicken villin headpiece) protein by using a simplified hydropathy scale. Special attention has been paid to the statics and dynamics of contact formation among the hydrophobic residues. The results obtained from this simple model appear to be surprisingly similar to several features observed in the folding of real proteins. The account concludes with a discussion of future problems.

Ternary copper(II) complexes of thiosemicarbazones and heterocyclic bases showing N3OS coordination as models for the type-2 centers of copper monooxygenases

Six ternary copper(II) complexes of general formulation [CuLB] (1-6), where L is dianionic ONS-donor thiosemicarbazones derived from the condensation of salicylaldehyde with thiosemicarbazides and B is NN-donor heterocyclic bases like 2,2'-bipyridine, 1,10-phenanthroline and 2,9-dimethyl-1,10-phenanthroline, are prepared from a reaction of copper(II) acetate hydrate with the heterocyclic base (B) and the thiosemicarbazone (H2L) in MeOH, and structurally characterized by X-ray diffraction technique. Crystal structures of the complexes display a distorted square-pyramidal (4 + 1) coordination geometry having the ONS-donor thiosemicarbazone bonded at the basal plane. The chelating heterocyclic bases exhibit axial-equatorial mode of bonding. The complexes are one-electron paramagnetic and they show axial X-band EPR spectra in DMF-toluene glass at 77 K giving g(parallel to)(A(parallel to)) and g(perpendicular to) values of similar to2.2 (175 x 10(-4) cm(-1)) and similar to2.0 indicating a {d(x2-y2)}(1) ground state. The complexes show a d-d band near 570 nm and a charge transfer band near 400 nm in DMF. The complexes are redox active and exhibit a quasireversible Cu(II)-Cu(I) couple in DMF-0.1 M tetrabutylammonium perchlorate near 0.1 V vs. SCE. They are catalytically active in the oxidation of ascorbic acid in presence of dioxygen. The complexes with a CuN3OS coordination model the ascorbate oxidation property of dopamine beta-hydroxylase and peptidylglycine a-hydroxylating monooxygeanase. (C) 2003 Elsevier Science B.V. All rights reserved.

Performance Evaluation of Distance-Hop Proportionality on Geometric Graph Models of Dense Sensor Networks

Wireless sensor networks can often be viewed in terms of a uniform deployment of a large number of nodes on a region in Euclidean space, e.g., the unit square. After deployment, the nodes self-organise into a mesh topology. In a dense, homogeneous deployment, a frequently used approximation is to take the hop distance between nodes to be proportional to the Euclidean distance between them. In this paper, we analyse the performance of this approximation. We show that nodes with a certain hop distance from a fixed anchor node lie within a certain annulus with probability approach- ing unity as the number of nodes n → ∞. We take a uniform, i.i.d. deployment of n nodes on a unit square, and consider the geometric graph on these nodes with radius r(n) = c q ln n n . We show that, for a given hop distance h of a node from a fixed anchor on the unit square,the Euclidean distance lies within [(1−ǫ)(h−1)r(n), hr(n)],for ǫ > 0, with probability approaching unity as n → ∞.This result shows that it is more likely to expect a node, with hop distance h from the anchor, to lie within this an- nulus centred at the anchor location, and of width roughly r(n), rather than close to a circle whose radius is exactly proportional to h. We show that if the radius r of the ge- ometric graph is fixed, the convergence of the probability is exponentially fast. Similar results hold for a randomised lattice deployment. We provide simulation results that il- lustrate the theory, and serve to show how large n needs to be for the asymptotics to be useful.

Voltage and Temperature Scalable Standard Cell Leakage Models Based on Stacks for Statistical Leakage Characterization

With extensive use of dynamic voltage scaling (DVS) there is increasing need for voltage scalable models. Similarly, leakage being very sensitive to temperature motivates the need for a temperature scalable model as well. We characterize standard cell libraries for statistical leakage analysis based on models for transistor stacks. Modeling stacks has the advantage of using a single model across many gates there by reducing the number of models that need to be characterized. Our experiments on 15 different gates show that we needed only 23 models to predict the leakage across 126 input vector combinations. We investigate the use of neural networks for the combined PVT model, for the stacks, which can capture the effect of inter die, intra gate variations, supply voltage(0.6-1.2 V) and temperature (0 - 100degC) on leakage. Results show that neural network based stack models can predict the PDF of leakage current across supply voltage and temperature accurately with the average error in mean being less than 2% and that in standard deviation being less than 5% across a range of voltage, temperature.

Voltage and Temperature Scalable Gate Delay and Slew Models Including Intra-Gate Variations

We investigate the feasibility of developing a comprehensive gate delay and slew models which incorporates output load, input edge slew, supply voltage, temperature, global process variations and local process variations all in the same model. We find that the standard polynomial models cannot handle such a large heterogeneous set of input variables. We instead use neural networks, which are well known for their ability to approximate any arbitrary continuous function. Our initial experiments with a small subset of standard cell gates of an industrial 65 nm library show promising results with error in mean less than 1%, error in standard deviation less than 3% and maximum error less than 11% as compared to SPICE for models covering 0.9- 1.1 V of supply, -40degC to 125degC of temperature, load, slew and global and local process parameters. Enhancing the conventional libraries to be voltage and temperature scalable with similar accuracy requires on an average 4x more SPICE characterization runs.

Voltage and Temperature Scalable Gate Delay and Slew Models Including Intra-Gate Variations

We investigate the feasibility of developing a comprehensive gate delay and slew models which incorporates output load, input edge slew, supply voltage, temperature, global process variations and local process variations all in the same model. We find that the standard polynomial models cannot handle such a large heterogeneous set of input variables. We instead use neural networks, which are well known for their ability to approximate any arbitrary continuous function. Our initial experiments with a small subset of standard cell gates of an industrial 65 nm library show promising results with error in mean less than 1%, error in standard deviation less than 3% and maximum error less than 11% as compared to SPICE for models covering 0.9- 1.1 V of supply, -40degC to 125degC of temperature, load, slew and global and local process parameters. Enhancing the conventional libraries to be voltage and temperature scalable with similar accuracy requires on an average 4x more SPICE characterization runs.

A comparative study of ordinary kriging and support vector machine models for the spatial variability of rock depth in Bangalore

In this paper, reduced level of rock at Bangalore, India is arrived from the 652 boreholes data in the area covering 220 sq.km. In the context of prediction of reduced level of rock in the subsurface of Bangalore and to study the spatial variability of the rock depth, ordinary kriging and Support Vector Machine (SVM) models have been developed. In ordinary kriging, the knowledge of the semivariogram of the reduced level of rock from 652 points in Bangalore is used to predict the reduced level of rock at any point in the subsurface of Bangalore, where field measurements are not available. A cross validation (Q1 and Q2) analysis is also done for the developed ordinary kriging model. The SVM is a novel type of learning machine based on statistical learning theory, uses regression technique by introducing e-insensitive loss function has been used to predict the reduced level of rock from a large set of data. A comparison between ordinary kriging and SVM model demonstrates that the SVM is superior to ordinary kriging in predicting rock depth.