Wenyu Qu¹, Zhiyang Li^2,*, Junfeng Wu², Yinan Wu³, Zhaobin Liu²

Computer Systems Science and Engineering, Vol.33, No.2, pp. 115-123, 2018, DOI:10.32604/csse.2018.33.115

Abstract Conventional image skeletonization techniques implicitly assume the pixel level connectivity. However, noise inside the object regions destroys the connectivity and exhibits sparseness in the image. We present a skeletonization algorithm designed for these kinds of sparse shapes. The skeletons are produced quickly by using three operations. First, initial skeleton nodes are selected by farthest point sampling with circles containing the maximum effective information. A skeleton graph of these nodes is imposed via inheriting the neighborhood of their associated pixels, followed by an edge collapse operation. Then a skeleton tting process based on feature-preserving Laplacian smoothing More >

Yuhong Zhang^1,*, Qinqin Wang¹, Yuling Li¹, Xindong Wu²

CMC-Computers, Materials & Continua, Vol.56, No.2, pp. 285-297, 2018, DOI:10.3970/cmc.2018.02604

Abstract Recently, the effectiveness of neural networks, especially convolutional neural networks, has been validated in the field of natural language processing, in which, sentiment classification for online reviews is an important and challenging task. Existing convolutional neural networks extract important features of sentences without local features or the feature sequence. Thus, these models do not perform well, especially for transition sentences. To this end, we propose a Piecewise Pooling Convolutional Neural Network (PPCNN) for sentiment classification. Firstly, with a sentence presented by word vectors, convolution operation is introduced to obtain the convolution feature map vectors. Secondly, More >

Daojian Zeng^1,3, Yuan Dai^1,3, Feng Li^1,3, R. Simon Sherratt², Jin Wang^3,*

CMC-Computers, Materials & Continua, Vol.55, No.1, pp. 121-136, 2018, DOI:10.3970/cmc.2018.055.121

Abstract Recently, many researchers have concentrated on using neural networks to learn features for Distant Supervised Relation Extraction (DSRE). These approaches generally use a softmax classifier with cross-entropy loss, which inevitably brings the noise of artificial class NA into classification process. To address the shortcoming, the classifier with ranking loss is employed to DSRE. Uniformly randomly selecting a relation or heuristically selecting the highest score among all incorrect relations are two common methods for generating a negative class in the ranking loss function. However, the majority of the generated negative class can be easily discriminated from… More >

M. Ohsaki¹, T. Miyamura², J. Y. Zhang³

CMES-Computer Modeling in Engineering & Sciences, Vol.111, No.4, pp. 303-333, 2016, DOI:10.3970/cmes.2016.111.303

Abstract A simple constitutive model, called semi-implicit model, for cyclic loading is proposed for steel materials used for structures such as building frames in civil engineering. The constitutive model is implemented in the E-Simulator, which is a software package for large-scale seismic response analysis. The constitutive relation is defined in an algorithmic manner based on the piecewise linear combined isotropic-kinematic hardening. Different rules are used for the first and subsequent loading states to incorporate characteristics such as yield plateau and Bauschinger effect of rolled mild steel materials. An optimization method is also presented for parameter identification More >

Taku Itoh¹, Susumu Nakata²

CMES-Computer Modeling in Engineering & Sciences, Vol.107, No.3, pp. 187-199, 2015, DOI:10.3970/cmes.2015.107.187

Abstract To speed up generating a scalar field g(x) based on a piecewise polynomial, a new method for determining field values that are indispensable to generate g(x) has been proposed. In the proposed method, an intermediate for generating g(x) does not required, i.e., the field values can directly be determined from given point data. Numerical experiments show that the computation time for determining the field values by the proposed method is about 10.4–12.7 times less than that of the conventional method. In addition, on the given points, the accuracy of g(x) obtained by using the proposed More >

H. Zheng¹, X. Peng^1,2,3, N. Hu^1,4

CMES-Computer Modeling in Engineering & Sciences, Vol.86, No.6, pp. 505-532, 2012, DOI:10.3970/cmes.2012.086.505

Abstract The static and kinematic shakedown of a functionally graded plate (FGP) is analyzed. The FGP is subjected coupled constant mechanical load and cyclically varying temperature. The FGP is composed of elastoplastic matrix and elastic particles, with the particle volume fraction varying along its thickness. The thermal and mechanical properties and their distributions are evaluated with a mean filed approach, which is based on the Eshelby's inclusion theory and takes into account directly the interaction between particles. The FGP is assumed to be separated into a number of thin layers, the thermal and mechanical properties in… More >

Xiao Zhao¹, Haitian Yang^1,2, Qiang Gao¹

CMES-Computer Modeling in Engineering & Sciences, Vol.74, No.2, pp. 139-160, 2011, DOI:10.3970/cmes.2011.074.139

Abstract A piecewised adaptive algorithm in the time domain is presented to solve the transient convection-diffusion heat transfer problem. By expanding all variables at a time interval, an initial and boundary value problem is decoupled into a series of recursive boundary value problems which can be solved by FEM or other well developed numerical schemes to deal with boundary value problems. A steady computing accuracy can be adaptively maintained via the power increase of the expansion, particularly when the step size varies in the whole computing process. Additionally for the nonlinear cases, there is no requirement More >

A. Colombo¹, U. Galvanetto²

CMES-Computer Modeling in Engineering & Sciences, Vol.53, No.3, pp. 235-254, 2009, DOI:10.3970/cmes.2009.053.235

Abstract The paper addresses the problem of computing the stable manifolds of equilibria and limit cycles of saddle type in piecewise smooth dynamical systems. All singular points that are generically present along one-dimensional or two-dimensional manifolds are classified and such a classification is then used to define a method for the numerical computation of the stable manifolds. Finally the proposed method is applied to the case of a stick-slip oscillator. More >

Briti Sundar Deb¹, Srinivasan Natesan²

CMES-Computer Modeling in Engineering & Sciences, Vol.38, No.2, pp. 179-200, 2008, DOI:10.3970/cmes.2008.038.179

Abstract This paper presents an almost second--order uniformly convergent Richardson extrapolation method for convection- dominated coupled system of boundary value problems. First, we solve the system by using the classical finite difference scheme on the layer resolving Shishkin mesh, and then we construct the Richardson approximation solution using the solutions obtained on N and 2N mesh intervals. Second-order parameter--uniform error estimate is derived. The proposed method is applied to a test example for verification of the theoretical results for the case ε ≤ N⁻¹. More >

G. Cocchetti², G. Maier², X. P. Shen³

CMES-Computer Modeling in Engineering & Sciences, Vol.3, No.3, pp. 279-298, 2002, DOI:10.3970/cmes.2002.003.279

Abstract Interface models mean here relationships between displacement jumps and tractions across a locus of displacement discontinuities. Frictional contact and quasi-brittle fracture interpreted by cohesive crack models are typical mechanical situations concerned by the present unifying approach. Plastic-softening multidissipative interface models are studied in piecewise linear formats, i.e. assuming linearity for yield functions, plastic potentials and relationships between static and kinematic internal variables. The properties and the pros and cons of such simplified models in a variety of formulations (fully non-holonomic in rates, holonomic and in finite steps), all mathematically described as linear complementarity problems, are More >