Efficient algorithms for learning kernels from multiple similarity matrices with general convex loss functions

In this paper we consider the problem of learning an n × n kernel matrix from m(1) similarity matrices under general convex loss. Past research have extensively studied the m = 1 case and have derived several algorithms which require sophisticated techniques like ACCP, SOCP, etc. The existing algorithms do not apply if one uses arbitrary losses and often can not handle m > 1 case. We present several provably convergent iterative algorithms, where each iteration requires either an SVM or a Multiple Kernel Learning (MKL) solver for m > 1 case. One of the major contributions of the paper is to extend the well knownMirror Descent(MD) framework to handle Cartesian product of psd matrices. This novel extension leads to an algorithm, called EMKL, which solves the problem in O(m2 log n 2) iterations; in each iteration one solves an MKL involving m kernels and m eigen-decomposition of n × n matrices. By suitably defining a restriction on the objective function, a faster version of EMKL is proposed, called REKL,which avoids the eigen-decomposition. An alternative to both EMKL and REKL is also suggested which requires only an SVMsolver. Experimental results on real world protein data set involving several similarity matrices illustrate the efficacy of the proposed algorithms.

Cross-polarization from the dipolar reservoir under the mismatched Hartmann-Hahn condition

Cross-polarization from the dipolar reservoir for a range of mismatched Hartmann-Hahn conditions has been considered. Experiment, in general, agrees with the dispersive Lorentzian behavior expected on the basis of quasi-equilibrium theory. It is observed that inclusion of additional mechanisms of polarization transfer lead to an improvment of the fit of the experimental results. The utility of extending the technique to the case of ordered long chain molecules, such as liquid crystals, for the measurement of the local dipolar field is also presented. (C) 2002 Elsevier Science (USA).

Dynamics of crystal size distributions with size-dependent rates

The growth and dissolution dynamics of nonequilibrium crystal size distributions (CSDs) can be determined by solving the governing population balance equations (PBEs) representing reversible addition or dissociation. New PBEs are considered that intrinsically incorporate growth dispersion and yield complete CSDs. We present two approaches to solving the PBEs, a moment method and a numerical scheme. The results of the numerical scheme agree with the moment technique, which can be solved exactly when powers on mass-dependent growth and dissolution rate coefficients are either zero or one. The numerical scheme is more general and can be applied when the powers of the rate coefficients are non-integers or greater than unity. The influence of the size dependent rates on the time variation of the CSDs indicates that as equilibrium is approached, the CSDs become narrow when the exponent on the growth rate is less than the exponent on the dissolution rate. If the exponent on the growth rate is greater than the exponent on the dissolution rate, then the polydispersity continues to broaden. The computation method applies for crystals large enough that interfacial stability issues, such as ripening, can be neglected. (C) 2002 Elsevier Science B.V. All rights reserved.

Indian Ocean dipole mode events in an ocean general circulation model

The evolution of the dipole mode (DM) events in the Indian Ocean is examined using an ocean model that is driven by the NCEP fluxes for the period 1975-1998. The positive DM events during 1997, 1994 and 1982 and negative DM events during 1996 and 1984-1985 are captured by the model and it reproduces both the surface and subsurface features associated with these events. In its positive phase, the DM is characterized by warmer than normal SST in the western Indian Ocean and cooler than normal SST in the eastern Indian Ocean. The DM events are accompanied by easterly wind anomalies along the equatorial Indian Ocean and upwelling-favorable alongshore wind anomalies along the coast of Sumatra. The Wyrtki jets are weak during positive DM events, and the thermocline is shallower than normal in the eastern Indian Ocean and deeper in the west. This anomaly pattern reverses during negative DM events. During the positive phase of the DM easterly wind anomalies excite an upwelling equatorial Kelvin wave. This Kelvin wave reflects from the eastern boundary as an upwelling Rossby wave which propagates westward across the equatorial Indian Ocean. The anomalies in the eastern Indian Ocean weaken after the Rossby wave passes. A similar process excites a downwelling Rossby wave during the negative phase. This Rossby wave is much weaker but wind forcing in the central equatorial Indian Ocean amplifies the downwelling and increases its westward phase speed. This Rossby wave initiates the deepening of the thermocline in the western Indian Ocean during the following positive phase of the DM. Rossby wave generated in the southern tropical Indian Ocean by Ekman pumping contributes to this warming. Concurrently, the temperature equation of the model shows upwelling and downwelling to be the most important mechanism during both positive events of 1994 and 1997. (C) 2002 Elsevier Science Ltd. All rights reserved.

Numerical modeling of heat and mass transfer in laser surface alloying: Non-equilibrium solidification effects

A transient macroscopic model is developed for studying heat and mass transfer in a single-pass laser surface alloying process, with particular emphasis on non-equilibrium solidification considerations. The solution for species concentration distribution requires suitable treatment of non-equilibrium mass transfer conditions. In this context, microscopic features pertaining to non-equilibrium effects on account of solutal undercooling are incorporated through the formulation of a modified partition-coefficient. The effective partition-coefficient is numerically modeled by Means of a number of macroscopically observable parameters related to the solidifying domain. The numerical model is so developed that the modifications on account of non-equilibrium solidification considerations can be conveniently implemented in existing numerical codes based on equilibrium solidification considerations.

Scaling analysis of momentum, heat, and mass transfer in binary alloy solidification problems

A systematic approach is developed for scaling analysis of momentum, heat and species conservation equations pertaining to the case of solidification of a binary mixture. The problem formulation and description of boundary conditions are kept fairly general, so that a large class of problems can be addressed. Analysis of the momentum equations coupled with phase change considerations leads to the establishment of an advection velocity scale. Analysis of the energy equation leads to an estimation of the solid layer thickness. Different regimes corresponding to different dominant modes of transport are simultaneously identified. A comparative study involving several cases of possible thermal boundary conditions is also performed. Finally, a scaling analysis of the species conservation equation is carried out, revealing the effect of a non-equilibrium solidification model on solute segregation and species distribution. It is shown that non-equilibrium effects result in an enhanced macrosegregation compared with the case of an equilibrium model. For the sake of assessment of the scaling analysis, the predictions are validated against corresponding computational results.

Symmetry-breaking transitions in equilibrium shapes of coherent precipitates: Effect of elastic anisotropy and inhomogeneity

We examine the symmetry-breaking transitions in equilibrium shapes of coherent precipitates in two-dimensional (2-D) systems under a plane-strain condition with the principal misfit strain components epsilon(xx)*. and epsilon(yy)*. For systems with cubic elastic moduli, we first show all the shape transitions associated with different values of t = epsilon(yy)*/epsilon(xx)*. We also characterize each of these transitions, by studying its dependence on elastic anisotropy and inhomogeneity. For systems with dilatational misfit (t = 1) and those with pure shear misfit (t = -1), the transition is from an equiaxed shape to an elongated shape, resulting in a break in rotational symmetry. For systems with nondilatational misfit (-1 < t < 1; t not equal 0), the transition involves a break in mirror symmetries normal to the x- and y-axes. The transition is continuous in all cases, except when 0 < t < 1. For systems which allow an invariant line (-1 less than or equal to t < 0), the critical size increases with an increase in the particle stiffness. However, for systems which do not allow an invariant line (0 < t less than or equal to 1), the critical size first decreases, reaches a minimum, and then starts increasing with increasing particle stiffness; moreover, the transition is also forbidden when the particle stiffness is greater than a critical value.

General theories for the electrical transport properties of carbon nanotubes

We have shown that the general theories of metals and semiconductors can be employed to understand the diameter and voltage dependency of current through metallic and semiconducting carbon nanotubes, respectively. The current through a semiconducting multiwalled carbon nanotube (MWCNT) is associated with the energy gap that is different for different shells. The contribution of the outermost shell is larger as compared to the inner shells. The general theories can also explain the diameter dependency of maximum current through nanotubes. We have also compared the current carrying ability of a MWCNT and an array of the same diameter of single wall carbon nanotubes (SWCNTs) and found that MWCNTs are better suited and deserve further investigation for possible applications as interconnects.

Equilibrium and relaxation in turbulent wakes

In order to study the memory of the larger eddies in turbulent shear flow, experiments have been conducted on plane turbulent wakes undergoing transition from an initial (carefully prepared) equilibrium state to a different final one, as a result of a nearly impulsive pressure gradient. It is shown that under the conditions of the experiments the equations of motion possess self-preserving solutions in the sense of Townsend (1956), but the observed behaviour of the wake is appreciably different when the pressure gradient is not very small, as the flow goes through a slow relaxation process before reaching final equilibrium. Measurements of the Reynolds stresse show that the approach to a new equilibrium state is exponential, with a relaxation length of the order of 103 momentum thicknesses. It is suggested that a flow satisfying the conditions required by a self-preservation analysis will exhibit equilibrium only if the relaxation length is small compared with a characteristic streamwise length scale of the flow.

Turbulence measurements in an ‘equilibrium’ axisymmetric wall jet Journal

This paper reports measurements of turbulent quantities in an axisymmetric wall jet subjected to an adverse pressure gradient in a conical diffuser, in such a way that a suitably defined pressure-gradient parameter is everywhere small. Self-similarity is observed in the mean velocity profile, as well as the profiles of many turbulent quantities at sufficiently large distances from the injection slot. Autocorrelation measurements indicate that, in the region of turbulent production, the time scale of ν fluctuations is very much smaller than the time scale of u fluctuations. Based on the data on these time scales, a possible model is proposed for the Reynolds stress. One-dimensional energy spectra are obtained for the u, v and w components at several points in the wall jet. It is found that self-similarity is exhibited by the one-dimensional wavenumber spectrum of $\overline{q^2}(=\overline{u^2}+\overline{v^2}+\overline{w^2})$, if the half-width of the wall jet and the local mean velocity are used for forming the non-dimensional wavenumber. Both the autocorrelation curves and the spectra indicate the existence of periodicity in the flow. The rate of dissipation of turbulent energy is estimated from the $\overline{q^2}$ spectra, using a slightly modified version of a previously suggested method.

A General Route to Nanoparticle Thin Films and Coatings

Nanoparticles thin films have wide range of applications such as nanoelectronics, magnetic storage devices, SERS substrate fabrication, optical grating and antireflective coating. Present work describes a method to prepare large area nanoparticles thin film of the order of few square centimeters. Thin film deposition has been done successfully on a wide range of conducting as well as non conducting substrates such as carbon-coated copper grid, silicon, m-plane of alumina, glass and (100) plane of NaCl single crystal. SEM, TEM and AFM studies have been done for microstructural characterization of the thin films. A basic mechanism has been proposed towards the understanding of the deposition process.

Three-dimensional double-diffusive convection and macrosegregation during non-equilibrium solidification of binary mixtures

A three-dimensional transient mathematical model (following a fixed-grid enthalpy-based continuum formulation) is used to study the interaction of double-diffusive natural convection and non-equilibrium solidification of a binary mixture in a cubic enclosure cooled from a side. Investigations are carried out for two separate test systems, one corresponding to a typical model "metal-alloy analogue" system and other corresponding to a real metal-alloy system. Due to stronger effects of solutal buoyancy in actual metal-alloy systems than in corresponding analogues, the convective transport mechanisms for the two cases are quite different. However, in both cases, similar elements of three-dimensionality are observed in the curvature and spacing of the projected streamlines. As a result of three-dimensional convective flow patterns, a significant solute macrosegregation is observed across the transverse sections of the cavity, which cannot be captured by two-dimensional simulations. (C) 2003 Elsevier Science Ltd. All rights reserved.