A Predictive Perfomance Model for Superscalar Processors

Designing and optimizing high performance microprocessors is an increasingly difficult task due to the size and complexity of the processor design space, high cost of detailed simulation and several constraints that a processor design must satisfy. In this paper, we propose the use of empirical non-linear modeling techniques to assist processor architects in making design decisions and resolving complex trade-offs. We propose a procedure for building accurate non-linear models that consists of the following steps: (i) selection of a small set of representative design points spread across processor design space using latin hypercube sampling, (ii) obtaining performance measures at the selected design points using detailed simulation, (iii) building non-linear models for performance using the function approximation capabilities of radial basis function networks, and (iv) validating the models using an independently and randomly generated set of design points. We evaluate our model building procedure by constructing non-linear performance models for programs from the SPEC CPU2000 benchmark suite with a microarchitectural design space that consists of 9 key parameters. Our results show that the models, built using a relatively small number of simulations, achieve high prediction accuracy (only 2.8% error in CPI estimates on average) across a large processor design space. Our models can potentially replace detailed simulation for common tasks such as the analysis of key microarchitectural trends or searches for optimal processor design points.

Fault-Diagnostic System using Support Vector Machines for Power Transmission Systems

This paper presents an approach for identifying the faulted line section and fault location on transmission systems using support vector machines (SVMs) for diagnosis/post-fault analysis purpose. Power system disturbances are often caused by faults on transmission lines. When fault occurs on a transmission system, the protective relay detects the fault and initiates the tripping operation, which isolates the affected part from the rest of the power system. Based on the fault section identified, rapid and corrective restoration procedures can thus be taken to minimize the power interruption and limit the impact of outage on the system. The approach is particularly important for post-fault diagnosis of any mal-operation of relays following a disturbance in the neighboring line connected to the same substation. This may help in improving the fault monitoring/diagnosis process, thus assuring secure operation of the power systems. In this paper we compare SVMs with radial basis function neural networks (RBFNN) in data sets corresponding to different faults on a transmission system. Classification and regression accuracy is reported for both strategies. Studies on a practical 24-Bus equivalent EHV transmission system of the Indian Southern region is presented for indicating the improved generalization with the large margin classifiers in enhancing the efficacy of the chosen model.

Time-frequency analysis of human motion during rhythmic exercises

Biomechanical signals due to human movements during exercise are represented in time-frequency domain using Wigner Distribution Function (WDF). Analysis based on WDF reveals instantaneous spectral and power changes during a rhythmic exercise. Investigations were carried out on 11 healthy subjects who performed 5 cycles of sun salutation, with a body-mounted Inertial Measurement Unit (IMU) as a motion sensor. Variance of Instantaneous Frequency (I.F) and Instantaneous Power (I.P) for performance analysis of the subject is estimated using one-way ANOVA model. Results reveal that joint Time-Frequency analysis of biomechanical signals during motion facilitates a better understanding of grace and consistency during rhythmic exercise.

Transport processes in one component plasmas using Landau equation

A system of transport equations have been obtained for plasma of electrons and having a background of positive ions in the presence of an electric and magnetic field. The starting kinetic equation is the well-known Landau kinetic equation. The distribution function of the kinetic equation has been expanded in powers of generalized Hermite polynomials and following Grad, a consistent set of transport equations have been obtained. The expressions for viscosity and heat conductivity have been deduced from the transport equation.

Particle dynamics in the channel flow of a turbulent particle-gas suspension at high Stokes number. Part 2. Comparison of fluctuating force simulations and experiments

The particle and fluid velocity fluctuations in a turbulent gas-particle suspension are studied experimentally using two-dimensional particle image velocimetry with the objective of comparing the experiments with the predictions of fluctuating force simulations. Since the fluctuating force simulations employ force distributions which do not incorporate the modification of fluid turbulence due to the particles, it is of importance to quantify the turbulence modification in the experiments. For experiments carried out at a low volume fraction of 9.15 x 10(-5) (mass loading is 0.19), where the viscous relaxation time is small compared with the time between collisions, it is found that the gas-phase turbulence is not significantly modified by the presence of particles. Owing to this, quantitative agreement is obtained between the results of experiments and fluctuating force simulations for the mean velocity and the root mean square of the fluctuating velocity, provided that the polydispersity in the particle size is incorporated in the simulations. This is because the polydispersity results in a variation in the terminal velocity of the particles which could induce collisions and generate fluctuations; this mechanism is absent if all of the particles are of equal size. It is found that there is some variation in the particle mean velocity very close to the wall depending on the wall-collision model used in the simulations, and agreement with experiments is obtained only when the tangential wall-particle coefficient of restitution is 0.7. The mean particle velocity is in quantitative agreement for locations more than 10 wall units from the wall of the channel. However, there are systematic differences between the simulations and theory for the particle concentrations, possibly due to inadequate control over the particle feeding at the entrance. The particle velocity distributions are compared both at the centre of the channel and near the wall, and the shape of the distribution function near the wall obtained in experiments is accurately predicted by the simulations. At the centre, there is some discrepancy between simulations and experiment for the distribution of the fluctuating velocity in the flow direction, where the simulations predict a bi-modal distribution whereas only a single maximum is observed in the experiments, although both distributions are skewed towards negative fluctuating velocities. At a much higher particle mass loading of 1.7, where the time between collisions is smaller than the viscous relaxation time, there is a significant increase in the turbulent velocity fluctuations by similar to 1-2 orders of magnitude. Therefore, it becomes necessary to incorporate the modified fluid-phase intensity in the fluctuating force simulation; with this modification, the mean and mean-square fluctuating velocities are within 20-30% of the experimental values.

Effect of convection and viscoelasticity on spinodal decomposition in fluids

Current analytical work on the effect of convection and viscoelasticity on the early and late stages of spinodal decomposition is briefly described. In the early stages, the effect of viscoelastic stresses was analysed using a simple Maxwell model for the stress, which was incorporated in the Langevin equation for the momentum field. The viscoelastic stresses are found to enhance the rate of decomposition. In the late stages, the pattern formed depends on the relative composition of the two species. Droplet spinodal decomposition occurs when the concentration of one of the species is small. Convective transport does not have a significant effect on the growth of a single droplet, but it does result in an attractive interaction between non - Brownian droplets which could lead to coalescence. The effect of convective transport for the growth of random interfaces in a near symmetric quench was analysed using an 'area distribution function', which gives the distribution of surface area of the interface in curvature space. It was found that the curvature of the interface decreases proportional to t in the late stages of spinodal decomposition, and the surface area also decreases proportional to t.

A neural network based automatic generation control design through reinforcement learning

This paper presents the design and implementation of a learning controller for the Automatic Generation Control (AGC) in power systems based on a reinforcement learning (RL) framework. In contrast to the recent RL scheme for AGC proposed by us, the present method permits handling of power system variables such as Area Control Error (ACE) and deviations from scheduled frequency and tie-line flows as continuous variables. (In the earlier scheme, these variables have to be quantized into finitely many levels). The optimal control law is arrived at in the RL framework by making use of Q-learning strategy. Since the state variables are continuous, we propose the use of Radial Basis Function (RBF) neural networks to compute the Q-values for a given input state. Since, in this application we cannot provide training data appropriate for the standard supervised learning framework, a reinforcement learning algorithm is employed to train the RBF network. We also employ a novel exploration strategy, based on a Learning Automata algorithm,for generating training samples during Q-learning. The proposed scheme, in addition to being simple to implement, inherits all the attractive features of an RL scheme such as model independent design, flexibility in control objective specification, robustness etc. Two implementations of the proposed approach are presented. Through simulation studies the attractiveness of this approach is demonstrated.

Intelligent Approach for Fault Diagnosis in Power Transmission Systems Using Support Vector Machines

This paper presents an approach for identifying the faulted line section and fault location on transmission systems using support vector machines (SVMs) for diagnosis/post-fault analysis purpose. Power system disturbances are often caused by faults on transmission lines. When fault occurs on a transmission system, the protective relay detects the fault and initiates the tripping operation, which isolates the affected part from the rest of the power system. Based on the fault section identified, rapid and corrective restoration procedures can thus be taken to minimize the power interruption and limit the impact of outage on the system. The approach is particularly important for post-fault diagnosis of any mal-operation of relays following a disturbance in the neighboring line connected to the same substation. This may help in improving the fault monitoring/diagnosis process, thus assuring secure operation of the power systems. In this paper we compare SVMs with radial basis function neural networks (RBFNN) in data sets corresponding to different faults on a transmission system. Classification and regression accuracy is reported for both strategies. Studies on a practical 24-Bus equivalent EHV transmission system of the Indian Southern region is presented for indicating the improved generalization with the large margin classifiers in enhancing the efficacy of the chosen model.

LOCAL TWO-DIMENSIONAL PARTICLE-IN-CELL SIMULATIONS OF THE COLLISIONLESS MAGNETOROTATIONAL INSTABILITY

The magnetorotational instability (MRI) is a crucial mechanism of angular momentum transport in a variety of astrophysical accretion disks. In systems accreting at well below the Eddington rate, such as the central black hole in the Milky Way (Sgr A*), the plasma in the disk is essentially collisionless. We present a nonlinear study of the collisionless MRI using first-principles particle-in-cell plasma simulations. We focus on local two-dimensional (axisymmetric) simulations, deferring more realistic three-dimensional simulations to future work. For simulations with net vertical magnetic flux, the MRI continuously amplifies the magnetic field, B, until the Alfven velocity, v(A), is comparable to the speed of light, c (independent of the initial value of v(A)/c). This is consistent with the lack of saturation of MRI channel modes in analogous axisymmetric MHD simulations. The amplification of the magnetic field by the MRI generates a significant pressure anisotropy in the plasma (with the pressure perpendicular to B being larger than the parallel pressure). We find that this pressure anisotropy in turn excites mirror modes and that the volume-averaged pressure anisotropy remains near the threshold for mirror mode excitation. Particle energization is due to both reconnection and viscous heating associated with the pressure anisotropy. Reconnection produces a distinctive power-law component in the energy distribution function of the particles, indicating the likelihood of non-thermal ion and electron acceleration in collisionless accretion disks. This has important implications for interpreting the observed emission-from the radio to the gamma-rays-of systems such as Sgr A*.

Subject independent human action recognition using spatio-depth information and meta-cognitive RBF network

In this paper, we present a machine learning approach for subject independent human action recognition using depth camera, emphasizing the importance of depth in recognition of actions. The proposed approach uses the flow information of all 3 dimensions to classify an action. In our approach, we have obtained the 2-D optical flow and used it along with the depth image to obtain the depth flow (Z motion vectors). The obtained flow captures the dynamics of the actions in space time. Feature vectors are obtained by averaging the 3-D motion over a grid laid over the silhouette in a hierarchical fashion. These hierarchical fine to coarse windows capture the motion dynamics of the object at various scales. The extracted features are used to train a Meta-cognitive Radial Basis Function Network (McRBFN) that uses a Projection Based Learning (PBL) algorithm, referred to as PBL-McRBFN, henceforth. PBL-McRBFN begins with zero hidden neurons and builds the network based on the best human learning strategy, namely, self-regulated learning in a meta-cognitive environment. When a sample is used for learning, PBLMcRBFN uses the sample overlapping conditions, and a projection based learning algorithm to estimate the parameters of the network. The performance of PBL-McRBFN is compared to that of a Support Vector Machine (SVM) and Extreme Learning Machine (ELM) classifiers with representation of every person and action in the training and testing datasets. Performance study shows that PBL-McRBFN outperforms these classifiers in recognizing actions in 3-D. Further, a subject-independent study is conducted by leave-one-subject-out strategy and its generalization performance is tested. It is observed from the subject-independent study that McRBFN is capable of generalizing actions accurately. The performance of the proposed approach is benchmarked with Video Analytics Lab (VAL) dataset and Berkeley Multimodal Human Action Database (MHAD). (C) 2013 Elsevier Ltd. All rights reserved.

Temporal Envelope Fit of Transient Audio Signals

We address the problem of temporal envelope modeling for transient audio signals. We propose the Gamma distribution function (GDF) as a suitable candidate for modeling the envelope keeping in view some of its interesting properties such as asymmetry, causality, near-optimal time-bandwidth product, controllability of rise and decay, etc. The problem of finding the parameters of the GDF becomes a nonlinear regression problem. We overcome the hurdle by using a logarithmic envelope fit, which reduces the problem to one of linear regression. The logarithmic transformation also has the feature of dynamic range compression. Since temporal envelopes of audio signals are not uniformly distributed, in order to compute the amplitude, we investigate the importance of various loss functions for regression. Based on synthesized data experiments, wherein we have a ground truth, and real-world signals, we observe that the least-squares technique gives reasonably accurate amplitude estimates compared with other loss functions.

Modeling Time-Varying Aggregate Interference in Cognitive Radio Systems, and Application to Primary Exclusive Zone Design

Accurately characterizing the time-varying interference caused to the primary users is essential in ensuring a successful deployment of cognitive radios (CR). We show that the aggregate interference at the primary receiver (PU-Rx) from multiple, randomly located cognitive users (CUs) is well modeled as a shifted lognormal random process, which is more accurate than the lognormal and the Gaussian process models considered in the literature, even for a relatively dense deployment of CUs. It also compares favorably with the asymptotically exact stable and symmetric truncated stable distribution models, except at high CU densities. Our model accounts for the effect of imperfect spectrum sensing, which depends on path-loss, shadowing, and small-scale fading of the link from the primary transmitter to the CU; the interweave and underlay modes or CR operation, which determine the transmit powers of the CUs; and time-correlated shadowing and fading of the links from the CUs to the PU-Rx. It leads to expressions for the probability distribution function, level crossing rate, and average exceedance duration. The impact of cooperative spectrum sensing is also characterized. We validate the model by applying it to redesign the primary exclusive zone to account for the time-varying nature of interference.

Small-Angle Neutron Scattering Studies of Hemoglobin Confined Inside Silica Tubes of Varying Sizes

In addition to the chemical nature of the surface, the dimensions of the confining host exert a significant influence on confined protein structures; this results in immense biological implications, especially those concerning the enzymatic activities of the protein. This study probes the structure of hemoglobin (Hb), a model protein, confined inside silica tubes with pore diameters that vary by one order of magnitude (approximate to 20-200 nm). The effect of confinement on the protein structure is probed by comparison with the structure of the protein in solution. Small-angle neutron scattering (SANS), which provides information on protein tertiary and quaternary structures, is employed to study the influence of the tube pore diameter on the structure and configuration of the confined protein in detail. Confinement significantly influences the structural stability of Hb and the structure depends on the Si-tube pore diameter. The high radius of gyration (R-g) and polydispersity of Hb in the 20 nm diameter Si-tube indicates that Hb undergoes a significant amount of aggregation. However, for Si-tube diameters greater or equal to 100 nm, the R-g of Hb is found to be in very close proximity to that obtained from the protein data bank (PDB) reported structure (R-g of native Hb=23.8 angstrom). This strongly indicates that the protein has a preference for the more native-like non-aggregated state if confined inside tubes of diameter greater or equal to 100 nm. Further insight into the Hb structure is obtained from the distance distribution function, p(r), and ab initio models calculated from the SANS patterns. These also suggest that the Si-tube size is a key parameter for protein stability and structure.

Dense shallow granular flows

Simplified equations are derived for a granular flow in the `dense' limit where the volume fraction is close to that for dynamical arrest, and the `shallow' limit where the stream-wise length for flow development (L) is large compared with the cross-stream height (h). The mass and diameter of the particles are set equal to 1 in the analysis without loss of generality. In the dense limit, the equations are simplified by taking advantage of the power-law divergence of the pair distribution function chi proportional to (phi(ad) - phi)(-alpha), and a faster divergence of the derivativ rho(d chi/d rho) similar to (d chi/d phi), where rho and phi are the density and volume fraction, and phi(ad) is the volume fraction for arrested dynamics. When the height h is much larger than the conduction length, the energy equation reduces to an algebraic balance between the rates of production and dissipation of energy, and the stress is proportional to the square of the strain rate (Bagnold law). In the shallow limit, the stress reduces to a simplified Bagnold stress, where all components of the stress are proportional to (partial derivative u(x)/partial derivative y)(2), which is the cross-stream (y) derivative of the stream-wise (x) velocity. In the simplified equations for dense shallow flows, the inertial terms are neglected in the y momentum equation in the shallow limit because the are O(h/L) smaller than the divergence of the stress. The resulting model contains two equations, a mass conservation equations which reduces to a solenoidal condition on the velocity in the incompressible limit, and a stream-wise momentum equation which contains just one parameter B which is a combination of the Bagnold coefficients and their derivatives with respect to volume fraction. The leading-order dense shallow flow equations, as well as the first correction due to density variations, are analysed for two representative flows. The first is the development from a plug flow to a fully developed Bagnold profile for the flow down an inclined plane. The analysis shows that the flow development length is ((rho) over barh(3)/B) , where (rho) over bar is the mean density, and this length is numerically estimated from previous simulation results. The second example is the development of the boundary layer at the base of the flow when a plug flow (with a slip condition at the base) encounters a rough base, in the limit where the momentum boundary layer thickness is small compared with the flow height. Analytical solutions can be found only when the stream-wise velocity far from the surface varies as x(F), where x is the stream-wise distance from the start of the rough base and F is an exponent. The boundary layer thickness increases as (l(2)x)(1/3) for all values of F, where the length scale l = root 2B/(rho) over bar. The analysis reveals important differences between granular flows and the flows of Newtonian fluids. The Reynolds number (ratio of inertial and viscous terms) turns out to depend only on the layer height and Bagnold coefficients, and is independent of the flow velocity, because both the inertial terms in the conservation equations and the divergence of the stress depend on the square of the velocity/velocity gradients. The compressibility number (ratio of the variation in volume fraction and mean volume fraction) is independent of the flow velocity and layer height, and depends only on the volume fraction and Bagnold coefficients.

Electron transfer statistics and thermal fluctuations in molecular junctions

We derive analytical expressions for probability distribution function (PDF) for electron transport in a simple model of quantum junction in presence of thermal fluctuations. Our approach is based on the large deviation theory combined with the generating function method. For large number of electrons transferred, the PDF is found to decay exponentially in the tails with different rates due to applied bias. This asymmetry in the PDF is related to the fluctuation theorem. Statistics of fluctuations are analyzed in terms of the Fano factor. Thermal fluctuations play a quantitative role in determining the statistics of electron transfer; they tend to suppress the average current while enhancing the fluctuations in particle transfer. This gives rise to both bunching and antibunching phenomena as determined by the Fano factor. The thermal fluctuations and shot noise compete with each other and determine the net (effective) statistics of particle transfer. Exact analytical expression is obtained for delay time distribution. The optimal values of the delay time between successive electron transfers can be lowered below the corresponding shot noise values by tuning the thermal effects. (C) 2015 AIP Publishing LLC.