MAPS: midline analysis and propagation of segmentation

Scenic word images undergo degradations due to motion blur, uneven illumination, shadows and defocussing, which lead to difficulty in segmentation. As a result, the recognition results reported on the scenic word image datasets of ICDAR have been low. We introduce a novel technique, where we choose the middle row of the image as a sub-image and segment it first. Then, the labels from this segmented sub-image are used to propagate labels to other pixels in the image. This approach, which is unique and distinct from the existing methods, results in improved segmentation. Bayesian classification and Max-flow methods have been independently used for label propagation. This midline based approach limits the impact of degradations that happens to the image. The segmented text image is recognized using the trial version of Omnipage OCR. We have tested our method on ICDAR 2003 and ICDAR 2011 datasets. Our word recognition results of 64.5% and 71.6% are better than those of methods in the literature and also methods that competed in the Robust reading competition. Our method makes an implicit assumption that degradation is not present in the middle row.

Classification calibration dimension for general multiclass losses

We study consistency properties of surrogate loss functions for general multiclass classification problems, defined by a general loss matrix. We extend the notion of classification calibration, which has been studied for binary and multiclass 0-1 classification problems (and for certain other specific learning problems), to the general multiclass setting, and derive necessary and sufficient conditions for a surrogate loss to be classification calibrated with respect to a loss matrix in this setting. We then introduce the notion of \emph{classification calibration dimension} of a multiclass loss matrix, which measures the smallest `size' of a prediction space for which it is possible to design a convex surrogate that is classification calibrated with respect to the loss matrix. We derive both upper and lower bounds on this quantity, and use these results to analyze various loss matrices. In particular, as one application, we provide a different route from the recent result of Duchi et al.\ (2010) for analyzing the difficulty of designing `low-dimensional' convex surrogates that are consistent with respect to pairwise subset ranking losses. We anticipate the classification calibration dimension may prove to be a useful tool in the study and design of surrogate losses for general multiclass learning problems.

Solid-solid collapse transition in a two dimensional model molecular system

Solid-solid collapse transition in open framework structures is ubiquitous in nature. The real difficulty in understanding detailed microscopic aspects of such transitions in molecular systems arises from the interplay between different energy and length scales involved in molecular systems, often mediated through a solvent. In this work we employ Monte-Carlo simulation to study the collapse transition in a model molecular system interacting via both isotropic as well as anisotropic interactions having different length and energy scales. The model we use is known as Mercedes-Benz (MB), which, for a specific set of parameters, sustains two solid phases: honeycomb and oblique. In order to study the temperature induced collapse transition, we start with a metastable honeycomb solid and induce transition by increasing temperature. High density oblique solid so formed has two characteristic length scales corresponding to isotropic and anisotropic parts of interaction potential. Contrary to the common belief and classical nucleation theory, interestingly, we find linear strip-like nucleating clusters having significantly different order and average coordination number than the bulk stable phase. In the early stage of growth, the cluster grows as a linear strip, followed by branched and ring-like strips. The geometry of growing cluster is a consequence of the delicate balance between two types of interactions, which enables the dominance of stabilizing energy over destabilizing surface energy. The nucleus of stable oblique phase is wetted by intermediate order particles, which minimizes the surface free energy. In the case of pressure induced transition at low temperature the collapsed state is a disordered solid. The disordered solid phase has diverse local quasi-stable structures along with oblique-solid like domains. (C) 2013 AIP Publishing LLC.

Suppression of gravitational instabilities by dominant dark matter halo in low-surface-brightness galaxies

The low-surface-brightness galaxies are gas rich and yet have a low star formation rate; this is a well-known puzzle. The spiral features in these galaxies are weak and difficult to trace, although this aspect has not been studied much. These galaxies are known to be dominated by the dark matter halo from the innermost regions. Here, we do a stability analysis for the galactic disc of UGC 7321, a low-surface-brightness, superthin galaxy, for which the various observational input parameters are available. We show that the disc is stable against local, linear axisymmetric and non-axisymmetric perturbations. The Toomre Q parameter values are found to be large (>> 1) mainly due to the low disc surface density, and the high rotation velocity resulting due to the dominant dark matter halo, which could explain the observed low star formation rate. For the stars-alone case, the disc shows finite swing amplification but the addition of dark matter halo suppresses that amplification almost completely. Even the inclusion of the low-dispersion gas which constitutes a high disc mass fraction does not help in causing swing amplification. This can explain why these galaxies do not show strong spiral features. Thus, the dynamical effect of a halo that is dominant from inner regions can naturally explain why star formation and spiral features are largely suppressed in low-surface-brightness galaxies, making these different from the high-surface-brightness galaxies.

Analysis of composite single lap joints using numerical and experimental approach

Adhesives are widely used to execute the assembly of aerospace and automotive structures due to their ability to join dissimilar materials, reduced stress concentration, and improved fatigue resistance. The mechanical behavior of adhesive joints can be studied either using analytical models or by conducting mechanical tests. However, the complexity owing to multiple interfaces, layers with different properties, material and geometric nonlinearity and its three-dimensional nature combine to increase the difficulty in obtaining an overall system of governing equations to predict the joint behavior. On the other hand, experiments are often time consuming and expensive due to a number of parameters involved. Finite element analysis (FEA) is profoundly used in recent years to overcome these limitations. The work presented in this paper involves the finite element modeling and analysis of a composite single lap joint where the adhesive-adherend interface region was modeled using connector elements. The computed stresses were compared with the experimental stresses obtained using digital image correlation technique. The results showed an agreement. Further, the failure load predicted using FEA was found to be closer to the actual failure load obtained by mechanical tests.

Homeostasis of functional maps in active dendrites emerges in the absence of individual channelostasis

The maintenance of ion channel homeostasis, or channelostasis, is a complex puzzle in neurons with extensive dendritic arborization, encompassing a combinatorial diversity of proteins that encode these channels and their auxiliary subunits, their localization profiles, and associated signaling machinery. Despite this, neurons exhibit amazingly stereotypic, topographically continuous maps of several functional properties along their active dendritic arbor. Here, we asked whether the membrane composition of neurons, at the level of individual ion channels, is constrained by this structural requirement of sustaining several functional maps along the same topograph. We performed global sensitivity analysis on morphologically realistic conductance-based models of hippocampal pyramidal neurons that coexpressed six well-characterized functional maps along their trunk. We generated randomized models by varying 32 underlying parameters and constrained these models with quantitative experimental measurements from the soma and dendrites of hippocampal pyramidal neurons. Analyzing valid models that satisfied experimental constraints on all six functional maps, we found topographically analogous functional maps to emerge from disparate model parameters with weak pairwise correlations between parameters. Finally, we derived a methodology to assess the contribution of individual channel conductances to the various functional measurements, using virtual knockout simulations on the valid model population. We found that the virtual knockout of individual channels resulted in variable, measurement and location-specific impacts across the population. Our results suggest collective channelostasis as a mechanism behind the robust emergence of analogous functional maps and have significant ramifications for the localization and targeting of ion channels and enzymes that regulate neural coding and homeostasis.

Asynchronous Gossip for Averaging and Spectral Ranking

We consider two variants of the classical gossip algorithm. The first variant is a version of asynchronous stochastic approximation. We highlight a fundamental difficulty associated with the classical asynchronous gossip scheme, viz., that it may not converge to a desired average, and suggest an alternative scheme based on reinforcement learning that has guaranteed convergence to the desired average. We then discuss a potential application to a wireless network setting with simultaneous link activation constraints. The second variant is a gossip algorithm for distributed computation of the Perron-Frobenius eigenvector of a nonnegative matrix. While the first variant draws upon a reinforcement learning algorithm for an average cost controlled Markov decision problem, the second variant draws upon a reinforcement learning algorithm for risk-sensitive control. We then discuss potential applications of the second variant to ranking schemes, reputation networks, and principal component analysis.

De novo inference of protein function from coarse-grained dynamics

Inference of molecular function of proteins is the fundamental task in the quest for understanding cellular processes. The task is getting increasingly difficult with thousands of new proteins discovered each day. The difficulty arises primarily due to lack of high-throughput experimental technique for assessing protein molecular function, a lacunae that computational approaches are trying hard to fill. The latter too faces a major bottleneck in absence of clear evidence based on evolutionary information. Here we propose a de novo approach to annotate protein molecular function through structural dynamics match for a pair of segments from two dissimilar proteins, which may share even <10% sequence identity. To screen these matches, corresponding 1 mu s coarse-grained (CG) molecular dynamics trajectories were used to compute normalized root-mean-square-fluctuation graphs and select mobile segments, which were, thereafter, matched for all pairs using unweighted three-dimensional autocorrelation vectors. Our in-house custom-built forcefield (FF), extensively validated against dynamics information obtained from experimental nuclear magnetic resonance data, was specifically used to generate the CG dynamics trajectories. The test for correspondence of dynamics-signature of protein segments and function revealed 87% true positive rate and 93.5% true negative rate, on a dataset of 60 experimentally validated proteins, including moonlighting proteins and those with novel functional motifs. A random test against 315 unique fold/function proteins for a negative test gave >99% true recall. A blind prediction on a novel protein appears consistent with additional evidences retrieved therein. This is the first proof-of-principle of generalized use of structural dynamics for inferring protein molecular function leveraging our custom-made CG FF, useful to all. (C) 2014 Wiley Periodicals, Inc.

A critical examination of the paradox of strength and ductility in ultrafine-grained metals

The paradox of strength and ductility is now well established and denotes the difficulty of simultaneously achieving both high strength and high ductility. This paradox was critically examined using a cast Al-7% Si alloy processed by high-pressure torsion (HPT) for up to 10 turns at a temperature of either 298 or 445 K. This processing reduces the grain size to a minimum of similar to 0.4 mu m and also decreases the average size of the Si particles. The results show that samples processed to high numbers of HPT turns exhibit both high strength and high ductility when tested at relatively low strain rates and the strain rate sensitivity under these conditions is similar to 0.14 which suggests that flow occurs by some limited grain boundary sliding and crystallographic slip. The results are also displayed on the traditional diagram for strength and ductility and they demonstrate the potential for achieving high strength and high ductility by increasing the number of turns in HPT.

A non-resonant mass sensor to eliminate the ``missing mass'' effect during mass measurement of biological materials

Resonant sensors and crystal oscillators for mass detection need to be excited at very high natural frequencies (MHz). Use of such systems to measure mass of biological materials affects the accuracy of mass measurement due to their viscous and/or viscoelastic properties. The measurement limitation of such sensor system is the difficulty in accounting for the ``missing mass'' of the biological specimen in question. A sensor system has been developed in this work, to be operated in the stiffness controlled region at very low frequencies as compared to its fundamental natural frequency. The resulting reduction in the sensitivity due to non-resonant mode of operation of this sensor is compensated by the high resolution of the sensor. The mass of different aged drosophila melanogaster (fruit fly) is measured. The difference in its mass measurement during resonant mode of operation is also presented. That, viscosity effects do not affect the working of this non-resonant mass sensor is clearly established by direct comparison. (C) 2014 AIP Publishing LLC.

Rollover-Preventive Force Synthesis at Active Suspensions in a Vehicle Performing a Severe Maneuver With Wheels Lifted Off

Among the intelligent safety technologies for road vehicles, active suspensions controlled by embedded computing elements for preventing rollover have received a lot of attention. The existing models for synthesizing and allocating forces in such suspensions are conservatively based on the constraints that are valid until no wheels lift off the ground. However, the fault tolerance of the rollover-preventive systems can be enhanced if the smart/active suspensions can intervene in the more severe situation in which the wheels have just lifted off the ground. The difficulty in computing control in the last situation is that the vehicle dynamics then passes into the regime that yields a model involving disjunctive constraints on the dynamics. Simulation of dynamics with disjunctive constraints in this context becomes necessary to estimate, synthesize, and allocate the intended hardware realizable forces in an active suspension. In this paper, we give an algorithm for the previously mentioned problem by solving it as a disjunctive dynamic optimization problem. Based on this, we synthesize and allocate the roll-stabilizing time-dependent active suspension forces in terms of sensor output data. We show that the forces obtained from disjunctive dynamics are comparable with existing force allocations and, hence, are possibly realizable in the existing hardware framework toward enhancing the safety and fault tolerance.

Bayesian parameter identification in dynamic state space models using modified measurement equations

When Markov chain Monte Carlo (MCMC) samplers are used in problems of system parameter identification, one would face computational difficulties in dealing with large amount of measurement data and (or) low levels of measurement noise. Such exigencies are likely to occur in problems of parameter identification in dynamical systems when amount of vibratory measurement data and number of parameters to be identified could be large. In such cases, the posterior probability density function of the system parameters tends to have regions of narrow supports and a finite length MCMC chain is unlikely to cover pertinent regions. The present study proposes strategies based on modification of measurement equations and subsequent corrections, to alleviate this difficulty. This involves artificial enhancement of measurement noise, assimilation of transformed packets of measurements, and a global iteration strategy to improve the choice of prior models. Illustrative examples cover laboratory studies on a time variant dynamical system and a bending-torsion coupled, geometrically non-linear building frame under earthquake support motions. (C) 2015 Elsevier Ltd. All rights reserved.

Kinematic flow patterns in slow deformation of a dense granular material

The kinematic flow pattern in slow deformation of a model dense granular medium is studied at high resolution using in situ imaging, coupled with particle tracking. The deformation configuration is indentation by a flat punch under macroscopic plane-strain conditions. Using a general analysis method, velocity gradients and deformation fields are obtained from the disordered grain arrangement, enabling flow characteristics to be quantified. The key observations are the formation of a stagnation zone, as in dilute granular flow past obstacles; occurrence of vortices in the flow immediately underneath the punch; and formation of distinct shear bands adjoining the stagnation zone. The transient and steady state stagnation zone geometry, as well as the strength of the vortices and strain rates in the shear bands, are obtained from the experimental data. All of these results are well-reproduced in exact-scale non-smooth contact dynamics simulations. Full 3D numerical particle positions from the simulations allow extraction of flow features that are extremely difficult to obtain from experiments. Three examples of these, namely material free surface evolution, deformation of a grain column below the punch and resolution of velocities inside the primary shear band, are highlighted. The variety of flow features observed in this model problem also illustrates the difficulty involved in formulating a complete micromechanical analytical description of the deformation.

Chemical Shifts to Metabolic Pathways: Identifying Metabolic Pathways Directly from a Single 2D NMR Spectrum

Identifying cellular processes in terms of metabolic pathways is one of the avowed goals of metabolomics studies. Currently, this is done after relevant metabolites are identified to allow their mapping onto specific pathways. This task is daunting due to the complex nature of cellular processes and the difficulty in establishing the identity of individual metabolites. We propose here a new method: ChemSMP (Chemical Shifts to Metabolic Pathways), which facilitates rapid analysis by identifying the active metabolic pathways directly from chemical shifts obtained from a single two-dimensional (2D) C-13-H-1] correlation NMR spectrum without the need for identification and assignment of individual metabolites. ChemSMP uses a novel indexing and scoring system comprised of a ``uniqueness score'' and a ``coverage score''. Our method is demonstrated on metabolic pathways data from the Small Molecule Pathway Database (SMPDB) and chemical shifts from the Human Metabolome Database (HMDB). Benchmarks show that ChemSMP has a positive prediction rate of >90% in the presence of deduttered data and can sustain the same at 60-70% even in the presence of noise, such as deletions of peaks and chemical shift deviations. The method tested on NMR data acquired for a mixture of 20 amino acids shows a success rate of 93% in correct recovery of pathways. When used on data obtained from the cell lysate of an unexplored oncogenic cell line, it revealed active metabolic pathways responsible for regulating energy homeostasis of cancer cells. Our unique tool is thus expected to significantly enhance analysis of NMIR-based metabolomics data by reducing existing impediments.