Class Prototype Research Articles

Zero-Shot Learning on Semantic Class Prototype Graph

Zero-Shot Learning (ZSL) for visual recognition is typically achieved by exploiting a semantic embedding space. In such a space, both seen and unseen class labels as well as image features can be embedded so that the similarity among them can be measured directly. In this work, we consider that the key to effective ZSL is to compute an optimal distance metric in the semantic embedding space. Existing ZSL works employ either euclidean or cosine distances. However, in a high-dimensional space where the projected class labels (prototypes) are sparse, these distances are suboptimal, resulting in a number of problems including hubness and domain shift. To overcome these problems, a novel manifold distance computed on a semantic class prototype graph is proposed which takes into account the rich intrinsic semantic structure, i.e., semantic manifold, of the class prototype distribution. To further alleviate the domain shift problem, a new regularisation term is introduced into a ranking loss based embedding model. Specifically, the ranking loss objective is regularised by unseen class prototypes to prevent the projected object features from being biased towards the seen prototypes. Extensive experiments on four benchmarks show that our method significantly outperforms the state-of-the-art.

A feasibility study of full-bridge type superconducting fault current controller on electric machine power stability

Recently, because of the advent of Smart Grid and integration of distributed generations, electrical power grids are facing uncountable challenges. Increase of fault current is one of such serious challenges and there are some fault current limiters (FCLs) that can limit the fault current. Existing grid protection FCLs, however, simply limit the fault current passively and can allow the existing protection coordination schemes to fail. This phenomenon leads to catastrophic failure in the complex system and may cause unpredictable power grid operation. Unlike a FCL, a superconducting fault current controller (SFCC) employs a full-bridge thyristor rectifier, a high temperature superconducting (HTS) DC reactor, and an embedded control unit to maintain the fault current level at a proper value by adjusting the phase angle of thyristors. This paper contains experimental and numerical analysis to design and fabricate a SFCC system for protection and stability improvement in power grids. At first, fundamental characteristics of a SFCC system were introduced. System circuit diagram and operational principles were proposed. Secondly, the developed small-scale SFCC system was introduced and verified. A 40 Vrms/30 Arms class prototype SFCC employing HTS DC reactor was fabricated and short circuit tests that simulate various fault conditions were implemented to verify the control performance of the fault current. Finally, the practical feasibility of application of the SFCC system to the power system was studied. The problems caused by three-phase faults from the power grid were surveyed and transient stability analysis of the power system was conducted by simulations. From the experimental and simulation results, we can verify the feasibility of the SFCC in power system.

Classification of complex Wishart matrices with a diffusion-reaction system guided by stochastic distances.

We propose a new method for polarimetric synthetic aperture radar (PolSAR) imagery classification based on stochastic distances in the space of random matrices obeying complex Wishart distributions. Given a collection of prototypes [Formula: see text] and a stochastic distance d(.,.), we classify any random matrix X using two criteria in an iterative set-up. First, we associate X with the class which minimizes the weighted stochastic distance w(m)d(X,Z(m)), where the positive weights w(m) are computed to max- imize the class discrimination power. Second, we improve the result by embedding the classification problem into a diffusion-reaction partial differential system where the diffusion term smooths the patches within the image, and the reaction term tends to move the pixel values towards the closest class prototype. In particular, the method inherits the benefits of speckle reduction by diffusion-like methods. Results on synthetic and real PolSAR data show the performance of the method.

A Commutation Overlap Period Reduction Method Expanding the Buck Operating Range for a Bidirectional Isolated DC–DC Converter

SUMMARYThis paper proposes a bidirectional isolated DC–DC converter with an active clamp circuit and two novel control methods in the buck mode. A circulating current reduction method decreases the conduction loss. A commutation overlap period reduction method maintains the output power by expanding the maximum actual duty cycle even if the input voltage decreases. The experimental results show that the proposed method increased the maximum output 1.5 times, without increasing the voltage applied to the low‐voltage side MOSFETs under the minimum input voltage condition. A 2‐kW class prototype showed a higher than 90% efficiency over a wide operating range for a greater than 20 input–output voltage ratio at 100‐kHz switching frequency.

Code Consistent Hashing Based on Information-Theoretic Criterion

Learning based hashing techniques have attracted broad research interests in the Big Media research area. They aim to learn compact binary codes which can preserve semantic similarity in the Hamming embedding. However, the discrete constraints imposed on binary codes typically make hashing optimizations very challenging. In this paper, we present a code consistent hashing ( CCH ) algorithm to learn discrete binary hash codes. To form a simple yet efficient hashing objective function, we introduce a new code consistency constraint to leverage discriminative information and propose to utilize the Hadamard code which favors an information-theoretic criterion as the class prototype. By keeping the discrete constraint and introducing an orthogonal constraint, our objective function can be minimized efficiently. Experimental results on three benchmark datasets demonstrate that the proposed CCH outperforms state-of-the-art hashing methods in both image retrieval and classification tasks, especially with short binary codes.

Particle acceleration and non-thermal emission in Pulsar Wind Nebulae from relativistic MHD simulations

Pulsar wind nebulae are among the most powerful particle accelerators in the Galaxy with acceleration efficiencies that reach up to 30% and maximum particle energies in the PeV range. In recent years relativistic axisymmetric MHD models have proven to be excellent tools for describing the physics of such objects, and particularly successful at explaining their high energy morphology, down to very fine details. Nevertheless, some important aspects of the physics of PWNe are still obscure: the mechanism(s) responsible for the acceleration of particles of all energies is (are) still unclear, and the origin of the lowest energy (radio emitting) particles is most mysterious. The correct interpretation of the origin of radio emitting particles is of fundamental importance, as this holds information about the amount of pair production in the pulsar magnetosphere, and hence on the role of pulsars as antimatter factories. On the other hand, the long lifetimes of these particles against synchrotron losses, allows them to travel far from their injection location, making their acceleration site difficult to constrain. As far as the highest energy (X and gamma-ray emitting) particles are concerned, their acceleration is commonly believed to occur at the pulsar wind termination shock. But since the upstream flow is thought to have non-uniform properties along the shock surface, important constraints on the acceleration mechanism(s) could come from exact knowledge of the location and flow properties where particles are being accelerated. We investigate in detail both topics by means of 2D numerical MHD simulations. Different assumptions on the origin of radio particles and more generally on the injection sites of all particles are considered, and the corresponding emission properties are computed. We discuss the physical constraints that can be inferred from comparison of the synthetic emission properties against multiwavelength observations of the PWN class prototype, the Crab Nebula.

Starting Characteristic Analysis of a Radial Inflow Turbine for the Regenerative Brayton Cycle

Microturbines have been developed as compact gas turbines to be applied in the regenerative Brayton cycle. A typical microturbine is composed of a centrifugal compressor and a radial inflow turbine. As such, the microturbine has a starting characteristic peculiar to radial inflow turbines. An idling state known as the windage point for mass flow rate can be formed because of improper inlet flow conditions for turbine expansion flow. The present study looked at the relationships between the radius ratio of the radial inflow turbine to the centrifugal compressor and the starting characteristic and at the effects of turbine inlet flow conditions on the starting characteristic. Fundamental equations for the relationships between the radius ratio and the starting characteristic were obtained. Effectiveness of the equations was compared with experiment results obtained with a 150 kW class prototype microturbine.

Multitarget fatty acid amide hydrolase/cyclooxygenase blockade suppresses intestinal inflammation and protects against nonsteroidal anti-inflammatory drug-dependent gastrointestinal damage.

The ability of nonsteroidal anti-inflammatory drugs (NSAIDs) to inhibit cyclooxygenase (Cox)-1 and Cox-2 underlies the therapeutic efficacy of these drugs, as well as their propensity to damage the gastrointestinal (GI) epithelium. This toxic action greatly limits the use of NSAIDs in inflammatory bowel disease (IBD) and other chronic pathologies. Fatty acid amide hydrolase (FAAH) degrades the endocannabinoid anandamide, which attenuates inflammation and promotes GI healing. Here, we describe the first class of systemically active agents that simultaneously inhibit FAAH, Cox-1, and Cox-2 with high potency and selectivity. The class prototype 4: (ARN2508) is potent at inhibiting FAAH, Cox-1, and Cox-2 (median inhibitory concentration: FAAH, 0.031 ± 0.002 µM; Cox-1, 0.012 ± 0.002 µM; and Cox-2, 0.43 ± 0.025 µM) but does not significantly interact with a panel of >100 off targets. After oral administration in mice, ARN2508 engages its intended targets and exerts profound therapeutic effects in models of intestinal inflammation. Unlike NSAIDs, ARN2508 causes no gastric damage and indeed protects the GI from NSAID-induced damage through a mechanism that requires FAAH inhibition. Multitarget FAAH/Cox blockade may provide a transformative approach to IBD and other pathologies in which FAAH and Cox are overactive.

CDFSIM: Efficient Stochastic Simulation Through Decomposition of Cumulative Distribution Functions of Transformed Spatial Patterns

Simulation of categorical and continuous variables is performed using a new pattern-based simulation method founded upon coding spatial patterns in one dimension. The method consists of, first, using a spatial template to extract information in the form of patterns from a training image. Patterns are grouped into a pattern database and, then, mapped to one dimension. Cumulative distribution functions of the one-dimensional patterns are built. Patterns are then classified by decomposing the cumulative distribution functions, and calculating class or cluster prototypes. During the simulation process, a conditioning data event is compared to the class prototype, and a pattern is randomly drawn from the best matched class. Several examples are presented so as to assess the performance of the proposed method, including conditional and unconditional simulations of categorical and continuous data sets. Results show that the proposed method is efficient and very well performing in both two and three dimensions. Comparison of the proposed method to the filtersim algorithm suggests that it is better at reproducing the multi-point configurations and main characteristics of the reference images, while less sensitive to the number of classes and spatial templates used in the simulations.

Alkylphosphocholines as Promising Antitumor Agents: Exploring the Role of Structural Features on the Hemolytic Potential.

Alkylphosphocholines (APC) are promising antitumor agents, which have the cellular membrane as primary target; however, red blood cell damage limits their wide therapeutic use. A variety of APC analogs has been synthesized and tested showing less hemolytic effect than the class prototype, Miltefosine (HePC). In this work, chemometric methods were applied to a set of 34 APC derivatives to identify the most relevant structural and molecular features of hemolytic activity. The APC derivatives were divided into three groups: (i) N-methylpiperidine and N-methylmorpholine derivatives with a long alkyl chain or flexible cyclopentadecyl rings, displaying a hemolytic rate of 17 %; (ii) adamantyl and cyclopentadecyl derivatives, showing an average hemolysis of 39 %; and, N,N,N-trimethylammonium, trans-N,N,N-trimethylcyclohexanamine, and trans-N,N,N-trimethylcyclopentanamine derivatives, whose average hemolysis was 41 %. The findings suggested that the presence of either bulky cationic head groups, or rings such as adamantyl and cyclohexyl, primarily increases the hemolysis of compounds with eleven atoms in the alkyl chain. Moreover, the macrocyclic cyclopentadecyl seems to be important to the hemolytic potential especially of compounds with five carbon atoms in the alkyl chain. Regarding linear carbon chain derivatives with no ring substitution, less bulky cationic head groups seem to favor hemolysis. Thus, in order to design more potent and less toxic APC antitumors, the reported structural/molecular patterns should not be included in their structure.

An independent limit on the axion mass from the variable white dwarf star R548

Pulsating white dwarfs with hydrogen-rich atmospheres, also known as DAV stars, can be used as astrophysical laboratories to constrain the properties of fundamental particles like axions. Comparing the measured cooling rates of these stars with the expected values from theoretical models allows us to search for sources of additional cooling due to the emission of weakly interacting particles. In this paper, we present an independent inference of the mass of the axion using the recent determination of the evolutionary cooling rate of R548, the DAV class prototype. We employ a state-of-the-art code which allows us to perform a detailed asteroseismological fit based on fully evolutionary sequences. Stellar cooling is the solely responsible of the rates of change of period with time () for the DAV class. Thus, the inclusion of axion emission in these sequences notably influences the evolutionary timescales, and also the expected pulsational properties of the DAV stars. This allows us to compare the theoretical values to the corresponding empirical rate of change of period with time of R548 to discern the presence of axion cooling. We found that if the dominant period at 213.13 s in R548 is associated with a pulsation mode trapped in the hydrogen envelope, our models indicate the existence of additional cooling in this pulsating white dwarf, consistent with axions of mass macos 2β ∼ 17.1 meV at a 2σ confidence level. This determination is in agreement with the value inferred from another well-studied DAV, G117-B15A. We now have two independent and consistent estimates of the mass of the axion obtained from DAVs, although additional studies of other pulsating white dwarfs are needed to confirm this value of the axion mass.

Many-to-many feature matching in object recognition: a review of three approaches

The mainstream object categorisation community relies heavily on object representations consisting of local image features, due to their ease of recovery and their attractive invariance properties. Object categorisation is therefore formulated as finding, that is, ‘detecting’, a one-to-one correspondence between image and model features. This assumption breaks down for categories in which two exemplars may not share a single local image feature. Even when objects are represented as more abstract image features, a collection of features at one scale (in one image) may correspond to a single feature at a coarser scale (in the second image). Effective object categorisation therefore requires the ability to match features many-to-many. In this paper, we review our progress on three independent object categorisation problems, each formulated as a graph matching problem and each solving the many-to-many graph matching problem in a different way. First, we explore the problem of learning a shape class prototype from a set of class exemplars which may not share a single local image feature. Next, we explore the problem of matching two graphs in which correspondence exists only at higher levels of abstraction, and describe a low-dimensional, spectral encoding of graph structure that captures the abstract shape of a graph. Finally, we embed graphs into geometric spaces, reducing the many-to-many graph-matching problem to a weighted point matching problem, for which efficient many-to-many matching algorithms exist.

Dimensional Reduction of Pattern-Based Simulation Using Wavelet Analysis

A pattern-based simulation technique using wavelet analysis is proposed for the simulation (wavesim) of categorical and continuous variables. Patterns are extracted by scanning a training image with a template and then storing them in a pattern database. The dimension reduction of patterns in the pattern database is performed by wavelet decomposition at certain scale and the approximate sub-band is used for pattern database classification. The pattern database classification is performed by the k-means clustering algorithm and classes are represented by a class prototype. For the simulation of categorical variables, the conditional cumulative density function (ccdf) for each class is generated based on the frequency of the individual categories at the central node of the template. During the simulation process, the similarity of the conditioning data event with the class prototypes is measured using the L 2-norm. When simulating categorical variables, the ccdf of the best matched class is used to draw a pattern from a class. When continuous variables are simulated, a random pattern is drawn from the best matched class. Several examples of conditional and unconditional simulation with two- and three- dimensional data sets show that the spatial continuity of geometric features and shapes is well reproduced. A comparative study with the filtersim algorithm shows that the wavesim performs better than filtersim in all examples. A full-field case study at the Olympic Dam base metals deposit, South Australia, simulates the lithological rock-type units as categorical variables. Results show that the proportions of various rock-type units in the hard data are well reproduced when similar to those in the training image; when rock-type proportions between the training image and hard data differ, the results show a compromise between the two.

A 10 kW Class Natural Gas-PEMFC Distributed Heat and Power Cogeneration System

A 10kW class prototype of natural gas-PEMFC distributed heat and power cogeneration system has been developed and demonstrated for the first time in China. In this system, fuel processing – including steps of natural gas desulfurisation, steam reforming, CO water gas shift, and CO preferential methanation for CO deep removal – has been applied. The 10kW class prototype was developed with design dimensions of 1600mm (L) × 1700mm (H) × 600mm (W). It mainly comprises natural gas reformer, reformate gas moisturiser, PEMFC stack, heat recovery unit, DC/AC inverter, and controlling computer. At a steady state, the reformate gas obtained through the fuel processor has a composition of higher than 75% H2, about 21% CO2, less than 4% CH4, less than 10 ppm(V/V) CO, and less than 10 ppb(g/V) sulfur in dry base. The prototype could supply about 8kW AC power and about 10kW heat. During increasing DC power output from 0.8kW to 8kW, DC and AC efficiency were higher than 40% and 35%, respectively, and the reformer efficiency higher than 75% (LHV). The efficiency of heat recovery reached 45%, so the total efficiency for cogeneration was higher than 80%. The efficiency of the inverter was basically higher than 90%. The PEMFC cogeneration system is suitable to be applied in small-scale hotels, community buildings etc. for supplying clean and high-efficiency electricity and hot water at the same time.

AT LAST—A V777 HER PULSATOR IN THE KEPLER FIELD

We present the discovery of the first-and so far the only-pulsating white dwarf star located in the field of view of the Kepler spacecraft. During our ongoing effort to search for compact pulsator candidates that can benefit from the near-continuous coverage of Kepler, we recently identified a faint DB star from spectroscopy obtained with the William Herschel Telescope. After establishing its physical parameters to be T{sub eff} = 24,950 K and log g = 7.91 dex, placing it right in the middle of the V777 Her instability strip, we immediately submitted the target for follow-up space observations. The Kepler light curve reveals a pulsation spectrum consisting of five modes that follow a sequence roughly equally spaced in period with a mean spacing of 37 s. The three strongest modes show a triplet structure with a mean splitting of 3.3 {mu}Hz. We conclude that this object is a V777 Her pulsator with a mass of {approx}0.56 M{sub sun}, and very similar to the class prototype.

Dynamic Norm-based Encoding for Unfamiliar Shapes in Human Visual Cortex

Previous studies have argued that faces and other objects are encoded in terms of their deviation from a class prototype or norm. This prototype is associated with a smaller neural population response compared with nonprototype objects. However, it is still unclear (1) whether a norm-based representation can emerge for unfamiliar or novel object classes through visual experience at the time scale of an experiment and (2) whether the results from previous studies are caused by the prototypicality of a stimulus, by the physical properties of individual stimuli independent from the stimulus distribution, and/or by the trial-to-trial adaptation. Here we show with a combined behavioral and event-related fMRI study in humans that a short amount of visual experience with exemplars from novel object classes determines which stimulus is represented as the norm. Prototypicality effects were observed at the behavioral level by behavioral asymmetries during a stimulus comparison task. The fMRI data revealed that class exemplars closest to the prototypes--the perceived average of each class--were associated with a smaller response in the anterior part of the visual object-selective cortex compared with other class exemplars. By dissociating between the physical characteristics and the prototypicality status of the stimuli and by controlling for trial-to-trial adaptation, we can firmly conclude for the first time that high-level visual areas represent the identity of exemplars using a dynamic, norm-based encoding principle.

Similarity-margin based feature selection for symbolic interval data

In this paper we propose a feature selection method for symbolic interval data based on similarity margin. In this method, classes are parameterized by an interval prototype based on an appropriate learning process. A similarity measure is defined in order to estimate the similarity between the interval feature value and each class prototype. Then, a similarity margin concept has been introduced. The heuristic search is avoided by optimizing an objective function to evaluate the importance (weight) of each interval feature in a similarity margin framework. The experimental results show that the proposed method selects meaningful features for interval data. In particular, the method we propose yields a significant improvement on classification task of three real-world datasets.

Learning in an Ambient Intelligent World: Enabling Technologies and Practices

The rapid evolution of information and communication technology opens a wide spectrum of opportunities to change our surroundings into an Ambient Intelligent (AmI) world. AmI is a vision of future information society, where people are surrounded by a digital environment that is sensitive to their needs, personalized to their requirements, anticipatory of their behavior, and responsive to their presence. It emphasizes on greater user friendliness, user empowerment, and more effective service support, with an aim to bring information and communication technology to everyone, every home, every business, and every school, thus improving the quality of human life. AmI unprecedentedly enhances learning experiences by endowing the users with the opportunities of learning in context, a breakthrough from the traditional education settings. In this survey paper, we examine some major characteristics of an AmI learning environment. To deliver a feasible and effective solution to ambient learning, we overview a few latest developed enabling technologies in context awareness and interactive learning. Associated practices are meanwhile reported. We also describe our experience in designing and implementing a smart class prototype, which allows teachers to simultaneously instruct both local and remote students in a context-aware and natural way.

Skeletal Shape Abstraction from Examples

Learning a class prototype from a set of exemplars is an important challenge facing researchers in object categorization. Although the problem is receiving growing interest, most approaches assume a one-to-one correspondence among local features, restricting their ability to learn true abstractions of a shape. In this paper, we present a new technique for learning an abstract shape prototype from a set of exemplars whose features are in many-to-many correspondence. Focusing on the domain of 2D shape, we represent a silhouette as a medial axis graph whose nodes correspond to "parts" defined by medial branches and whose edges connect adjacent parts. Given a pair of medial axis graphs, we establish a many-to-many correspondence between their nodes to find correspondences among articulating parts. Based on these correspondences, we recover the abstracted medial axis graph along with the positional and radial attributes associated with its nodes. We evaluate the abstracted prototypes in the context of a recognition task.

Some Symmetry Based Classifiers

In this paper, a novel point symmetry based pattern classifier (PSC) is proposed. A recently developed point symmetry based distance is utilized to determine the amount of point symmetry of a particular test pattern with respect to a class prototype. Kd-tree based nearest neighbor search is used for reducing the complexity of point symmetry distance computation. The proposed point symmetry based classifier is well-suited for classifying data sets having point symmetric classes, irrespective of any convexity, overlap or size. In order to classify data sets having line symmetry property, a line symmetry based classifier (LSC) along the lines of PSC is thereafter proposed in this paper. To measure the total amount of line symmetry of a particular point in a class, a new definition of line symmetry based distance is also provided. Proposed LSC preserves the advantages of PSC. The performance of PSC and LSC are demonstrated in classifying fourteen artificial and real-life data sets of varying complexities. For the purpose of comparison, k-NN classifier and the well-known support vector machine (SVM) based classifiers are executed on the data sets used here for the experiments. Statistical analysis, ANOVA, is also performed to compare the performance of these classification techniques.