記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.2299/jsp.23.41

researchmap

DOI： 10.1016/j.jcss.2018.10.006

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

DOI： 10.1109/ICCNC.2019.8685549

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

DOI： 10.1109/TENCON.2018.8650418

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.1109/LCSYS.2018.2843679

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：Springer New York LLC  

Multiplicative update rules are a well-known computational method for nonnegative matrix factorization. Depending on the error measure between two matrices, various types of multiplicative update rules have been proposed so far. However, their convergence properties are not fully understood. This paper provides a sufficient condition for a general multiplicative update rule to have the global convergence property in the sense that any sequence of solutions has at least one convergent subsequence and the limit of any convergent subsequence is a stationary point of the optimization problem. Using this condition, it is proved that many of the existing multiplicative update rules have the global convergence property if they are modified slightly so that all variables take positive values. This paper also proposes new multiplicative update rules based on Kullback–Leibler, Gamma, and Rényi divergences. It is shown that these three rules have the global convergence property if the same modification as above is made.

DOI： 10.1007/s10589-018-9997-y

Web of Science

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Institute of Electronics, Information and Communication, Engineers, IEICE  

Nonnegative Matrix Factorization (NMF) with sparseness and smoothness constraints has attracted increasing attention. When these properties are considered, NMF is usually formulated as an optimization problem in which a linear combination of an approximation error term and some regularization terms must be minimized under the constraint that the factor matrices are nonnegative. In this paper, we focus our attention on the error measure based on the Euclidean distance and propose a new iterative method for solving those optimization problems. The proposed method is based on the Hierarchical Alternating Least Squares (HALS) algorithm developed by Cichocki et al. We first present an example to show that the original HALS algorithm can increase the objective value. We then propose a new algorithm called the Gauss-Seidel HALS algorithm that decreases the objective value monotonically. We also prove that it has the global convergence property in the sense of Zangwill. We finally verify the effectiveness of the proposed algorithm through numerical experiments using synthetic and real data.

DOI： 10.1587/transfun.E100.A.2925

Web of Science

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：ELSEVIER SCIENCE BV  

This paper studies the problem of finding graphs that locally maximize the clustering coefficient in the space of graphs with a fixed degree sequence. Such a graph is characterized by the property that the clustering coefficient cannot be increased, no matter how a single 2-switch is applied. First, an explicit formula for the amount of change in the clustering coefficient of a graph caused by a single 2-switch is given. Next, some classes of graphs with the property stated above are presented. An example of such a graph is the one obtained from a tree by replacing its edges with cliques with the same order. (C) 2016 Elsevier B.V. All rights reserved.

DOI： 10.1016/j.dam.2016.10.002

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：Institute of Electrical and Electronics Engineers Inc.  

The second smallest eigenvalue of the Laplacian matrix, also known as the algebraic connectivity, characterizes the performance of some dynamic processes on networks, such as consensus in multiagent networks, synchronization of coupled oscillators, random walks on graphs, and so on. In a multiagent network, for example, the larger the algebraic connectivity of the graph representing interactions between agents is, the faster the convergence speed of a representative consensus algorithm is. This paper tackles the problem of finding graphs that maximize or locally maximize the algebraic connectivity in the space of graphs with a fixed number of vertices and edges. It is shown that some well-known classes of graphs such as star graphs, cycle graphs, complete bipartite graphs, and circulant graphs are algebraic connectivity maximizers or local maximizers under certain conditions.

DOI： 10.1109/TCNS.2015.2503561

Web of Science

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：Springer Verlag  

We propose a novel iterative algorithm for nonnegative matrix factorization with the alpha-divergence. The proposed algorithm is based on the coordinate descent and the Newton method. We show that the proposed algorithm has the global convergence property in the sense that the sequence of solutions has at least one convergent subsequence and the limit of any convergent subsequence is a stationary point of the corresponding optimization problem. We also show through numerical experiments that the proposed algorithm is much faster than the multiplicative update rule.

DOI： 10.1007/978-3-319-70087-8_36

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：European Signal Processing Conference, EUSIPCO  

Multiplicative updates are widely used for nonnegative matrix factorization (NMF) as an efficient computational method. In this paper, we consider a class of constrained optimization problems in which a polynomial function of the product of two matrices is minimized subject to the nonnegativity constraints. These problems are closely related to NMF because the polynomial function covers many error function used for NMF. We first derive a multiplicative update rule for those problems by using the unified method developed by Yang and Oja. We next prove that a modified version of the update rule has the global convergence property in the sense of Zangwill under certain conditions. This result can be applied to many existing multiplicative update rules for NMF to guarantee their global convergence.

DOI： 10.1109/EUSIPCO.2016.7760286

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Algebraic connectivity of a network, which is defined as the second smallest eigenvalue of the Laplacian matrix, represents how strongly the network is connected. This paper proposes a new decentralized discrete-time algorithm for the estimation of the algebraic connectivity of multiagent networks. The validity of the proposed algorithm is verified by theoretical analysis and numerical experiments.

DOI： 10.1109/APCCAS.2016.7803941

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Multiplicative updates are widely used for nonnegative matrix factorization (NMF) as an efficient computational method. In this paper, we consider a class of constrained optimization problems in which a polynomial function of the product of two matrices is minimized subject to the nonnegativity constraints. These problems are closely related to NMF because the polynomial function covers many error function used for NMF. We first derive a multiplicative update rule for those problems by using the unified method developed by Yang and Oja. We next prove that a modified version of the update rule has the global convergence property in the sense of Zangwill under certain conditions. This result can be applied to many existing multiplicative update rules for NMF to guarantee their global convergence.

DOI： 10.1109/EUSIPCO.2016.7760286

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Hierarchical alternating least squares (HALS) algorithms are efficient computational methods for nonnegative matrix factorization (NMF). Given an initial solution, HALS algorithms update the solution block by block iteratively so that the error decreases monotonically. However, update rules in HALS algorithms are not well-defined. In addition, due to this problem, the convergence of the sequence of solutions to a stationary point cannot be proved theoretically. In this paper, we consider the HALS algorithm for the Frobenius norm-based NMF, and prove that a modified version has the global convergence property in the sense of Zangwill.

DOI： 10.1109/CAMSAP.2015.7383726

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：SPRINGER  

Nonnegative matrix factorization (NMF) is the problem of approximating a given nonnegative matrix by the product of two nonnegative matrices. The multiplicative updates proposed by Lee and Seung are widely used as efficient computational methods for NMF. However, the global convergence of these updates is not formally guaranteed because they are not defined for all pairs of nonnegative matrices. In this paper, we consider slightly modified versions of the original multiplicative updates and study their global convergence properties. The only difference between the modified updates and the original ones is that the former do not allow variables to take values less than a user-specified positive constant. Using Zangwill's global convergence theorem, we prove that any sequence of solutions generated by either of those modified updates has at least one convergent subsequence and the limit of any convergent subsequence is a stationary point of the corresponding optimization problem. Furthermore, we propose algorithms based on the modified updates that always stop within a finite number of iterations.

DOI： 10.1007/s10589-013-9593-0

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：ELSEVIER SCIENCE BV  

A simple connected undirected graph G is called a clustering coefficient locally maximizing graph if its clustering coefficient is not less than that of any simple connected graph obtained from G-by-rewiring an edge, that is, removing an edge and adding a new edge. In this paper, we present some new classes of clustering coefficient locally maximizing graphs. We first show that any graph composed of multiple cliques with orders greater than two sharing one vertex is a clustering coefficient locally maximizing graph. We next show that any graph obtained from a tree by replacing edges with cliques with the same order other than four is a clustering coefficient locally maximizing graph. We also extend the latter result to a more general class. (C) 2013 Elsevier B.V. All rights reserved.

DOI： 10.1016/j.dam.2013.09.013

Web of Science

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

There have been proposed various types of multiplicative updates for nonnegative matrix factorization. However, these updates have a serious drawback in common: they are not defined for all pairs of nonnegative matrices. Furthermore, due to this drawback, their global convergence in the sense of Zangwill's theorem cannot be proved theoretically. In this paper, we consider slightly modified versions of various multiplicative update rules, that are defined for all pairs of matrices in the domain, and show that many of them have the boundedness property. This property is a necessary condition for update rules to be globally convergent in the sense of Zangwill's theorem. © 2013 IEEE.

DOI： 10.1109/CAMSAP.2013.6714055

Web of Science

Scopus

researchmap

DOI： 10.1587/nolta.4.430

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE COMPUTER SOC  

We elucidate the potential limit of single-frame super-resolution by information theory. Though various algorithms for super-resolution have been proposed, there exist only few works that evaluate the performance of super-resolution to our knowledge. Our key idea is that "single-frame super-resolution task can be regarded as channel coding in information theory." Based on this recognition, we can apply some techniques of information theory to the analysis of single-frame super-resolution. As its first step, we clarify the potential limit of single-frame super-resolution. For this purpose, we use a model of Yang et al. (2008) as a statistical model of natural images. As a result, we elucidate the condition that "arbitrary high-resolution natural image can be potentially recovered with arbitrarily small error by single-frame super-resolution." This condition depends on S/N ratio and blurring parameter. We investigate numerically whether this condition is satisfied or not for several situations.

DOI： 10.1109/EST.2012.32

Web of Science

researchmap

DOI： 10.1587/nolta.3.301

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.1587/nolta.2.443

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：SPRINGER-VERLAG BERLIN  

Nonnegative matrix factorization (NMF) is to approximate a given large nonnegative matrix by the product of two small nonnegative matrices. Although the multiplicative update algorithm is widely used as an efficient computation method for NMF, it has a serious drawback that the update formulas are not well-defined because they are expressed in the form of a fraction. Furthermore, due to this drawback, the global convergence of the algorithm has not been guaranteed. In this paper, we consider NMF in which the approximation error is measured by the Euclidean distance between two matrices. We propose a modified multiplicative update algorithm in order to overcome the drawback of the original version and prove its global convergence.

DOI： 10.1007/978-3-642-24958-7_76

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

DOI： 10.1587/nolta.2.432

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：PERGAMON-ELSEVIER SCIENCE LTD  

It is well known that one-dimensional cellular neural networks (1D CNNs) with the template A = [1, 2, -1] can perform connected component detection (CCD). However, this has been confirmed only by numerical and laboratory experiments. In this paper, sufficient conditions for 1D CNNs to perform CCD are obtained through theoretical analysis. Main result shows that a wide class of templates including A = [1, 2, -1] can be used for CCD. (C) 2010 Elsevier Ltd. All rights reserved.

DOI： 10.1016/j.nonrwa.2010.05.007

Web of Science

researchmap

researchmap

DOI： 10.1109/CNNA.2010.5430264

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

The problem of finding the minimum and maximum values of the clustering coefficients of graphs with the fixed numbers of vertices and edges is considered. It is shown that when a graph has n vertices and the average degree is n/2 the clustering coefficient can take values from 0 to 1 1/2(n 1).

DOI： 10.1109/ECCTD.2009.5275108

Web of Science

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

Stable patterns that can be realized by a class of 1-D two-layer cellular neural networks (CNNs) are studied in this paper. We first introduce the notions of potentially stable pattern, potentially stable local pattern, and local pattern set. We then show that all of 256 possible sets can be realized as the local pattern set of the two-layer CNN, while only 59 sets can be realized as the local pattern set of the single-layer CNN. We also propose a simple,way to optimize the template values of the CNN, which is formulated as a set of linear programming problems, and present the obtained values for all of 256 sets.

DOI： 10.1109/TCSI.2008.925828

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

Global convergence of the sequential minimal optimization (SMO) algorithm for support vector regression (SVR) is studied in this paper. Given l training samples, SVR is formulated as a convex quadratic programming (QP) problem with l pairs of variables. We prove that if two pairs of variables violating the optimality condition are chosen for update in each step and subproblems are solved in a certain way, then the SMO algorithm always stops within a finite number of iterations after finding an optimal solution. Also, efficient implementation techniques for the SMO algorithm are presented and compared experimentally with other SMO algorithms.

DOI： 10.1109/TNN.2007.915116

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Recently a sufficient condition for the recurrent neural network with the piecewise-linear output characteristic to generate a prescribed periodic sequence of binary vectors such that every two consecutive vectors differ in exactly one component has been derived. If a recurrent neural network satisfies this condition, it is guaranteed that any state trajectory of the network passes through the periodic sequence of regions corresponding to the periodic sequence of binary vectors. However, the asymptotic behavior of the state trajectories has not been clarified yet. In this paper, we study asymptotic behavior of state trajectories of recurrent neural networks satisfying the above-mentioned sufficient condition, and derive a criterion for state trajectories to converge a unique limit cycle.

DOI： 10.1109/IJCNN.2008.4633836

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

Stable patterns that can be realized by a class of 1-D two-layer cellular neural networks (CNNs) are studied in this paper. We first introduce the notions of potentially stable pattern, potentially stable local pattern, and local pattern set. We then show that all of 256 possible sets can be realized as the local pattern set of the two-layer CNN, while only 59 sets can be realized as the local pattern set of the single-layer CNN. We also propose a simple way to optimize the template values of the CNN, which is formulated as a set of linear programming problems, and present the obtained values for all of 256 sets. © 2008 IEEE.

DOI： 10.1109/TCSI.2008.925828

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE COMPUTER SOC  

A novel algorithm for multi-class support vector machines (SVMs) is proposed in this paper The tree constructed in our algorithm consists of a series of two-class SVMs. Considering both separability and balance, in each iteration multi-class patterns are divided into two sets according to the distances between pairwise classes and the number of patterns in each class. This algorithm can well treat with the unequally distributed problems. The efficiency of the proposed method are verified by the experimental results.

DOI： 10.1109/ICACTE.2008.48

Web of Science

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：SPRINGER-VERLAG BERLIN  

Decomposition method has been widely used to efficiently solve the large size quadratic programming (QP) problems arising in support vector regression (SVR). In a decomposition method, a large QP problem is decomposed into a series of smaller QP subproblems, which can be solved much faster than the original one. In this paper, we analyze the global convergence of decomposition methods for SVR. We will show the decomposition methods for the convex programming problem formulated by Flake and Lawrence always stop within a finite number of iterations.

DOI： 10.1007/978-3-540-87732-5_74

Web of Science

researchmap

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Connected component detection (CCD) is an important image processing task done by one-dimensional cellular neural networks (1-D CNNs). Recently, some sufficient conditions for 1-D CNNs with the antisymmetric template A = [s,p, -s] to perform CCD have been derived under the assumption that the outputs of the boundary cells are set to 1 or -1. In this paper, we extend these results to I-D CNNs with the opposite-sign template A = [r,p, -s]. It is shown that the 1-D CNN can perform CCD for a wide range of parameter space. Therefore we can design I-D CNNs which not only can perform CCD but also are robust against small perturbations of the parameters.

DOI： 10.1109/ISCAS.2007.378101

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

We give the necessary and sufficient conditions for a one-dimensional discrete-time autonomous binary cellular neural networks to be stable in the case of fixed boundary. The results are complete generalization of our previous one [16] in which the symmetrical connections were assumed. The conditions are compared with some stability conditions so far known.

DOI： 10.1093/ietfec/e89-a.12.3963

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

Decomposition methods are well-known techniques for solving quadratic programming (QP) problems arising in support vector machines (SVMs). In each iteration of a decomposition method, a small number of variables are selected and a QP problem with only the selected variables is solved. Since large matrix computations are not required,. decomposition methods are applicable to large QP problems. In this paper,. We will make a rigorous analysis of the global convergence of general decomposition methods for SVMs. We first introduce a relaxed version of the optimality condition for the QP problems and then prove that a decomposition method reaches a solution satisfying this relaxed optimality condition within a finite number of iterations under a very mild condition on how to select variables.

DOI： 10.1109/TNN.2006.880584

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

We give necessary and sufficient conditions for a I-D DBCNN (1-dimensional discrete-time binary cellular neural network) with an external input to be stable in terms of connection coefficients. The results are generalization of our previous one [18], [19] in which the input was assumed to be zero.

DOI： 10.1093/ietfec/e89-a.10.2825

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

A novel method to simplify decision functions of support vector machines (SVMs) is proposed in this paper. In our method, a decision function is determined first in a usual way by using all training samples. Next those support vectors which contribute less to the decision function are excluded from the training samples. Finally a new decision function is obtained by using the remaining samples. Experimental results show that the proposed method can effectively simplify decision functions of SVMs without reducing the generalization capability.

DOI： 10.1093/ietfec/e89-a.10.2795

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

We study global dynamical behavior of cellular neural networks (CNNs) consisting of two cells. Since the output characteristic of each cell is expressed by a piecewise-linear function, a CNN with two cells is considered as a planar piecewise-linear dynamical system. We present the necessary and sufficient condition for such a CNN to be completely stable under the assumptions that: 1) self-coupling coefficients take the same value greater than one and 2) biases are set to zero. The condition is explicitly expressed in terms of coupling coefficients between cells.

DOI： 10.1109/TCSII.2006.876466

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：SPRINGER-VERLAG BERLIN  

A novel sequential minimal optimization (SMO) algorithm for support vector regression is proposed. This algorithm is based on Flake and Lawrence's SMO in which convex optimization problems with iota variables are solved instead of standard quadratic programming problems with 2 iota variables where iota is the number of training samples, but the strategy for working set selection is quite different. Experimental results show that the proposed algorithm is much faster than Flake and Lawrence's SMO and comparable to the fastest conventional SMO.

DOI： 10.1007/11893028_92

Web of Science

researchmap

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

A sequential minimal optimization (SMO) algorithm for support vector regression (SVR) has recently been proposed by Flake and Lawrence. However, the convergence of their algorithm has not been proved so far. In this paper, we consider an SMO algorithm, which deals with the same optimization problem as Flake and Lawrence's SMO, and give a rigorous proof that it always stops within a finite number of iterations.

DOI： 10.1109/IJCNN.2006.246703

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Global dynamical behavior of one-dimensional cellular neural networks (1-D CNNs) with the antisymmetric template A = [s, p, -s] is studied in this paper. Under the assumption that the outputs of the boundary cells are fixed to 1 or -1, a new sufficient condition for such CNNs to perform connected component detection will be presented.

DOI： 10.1109/ISCAS.2006.1693048

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

Sequential minimal optimization (SMO) algorithm is one of the simplest decomposition methods for learning of support vector machines (SVMs). Keerthi and Gilbert have recently studied the convergence property of SMO algorithm and given a proof that SMO algorithm always stops within a finite number of iterations. In this letter, we point out the incompleteness of their proof and give a more rigorous proof.

DOI： 10.1109/TNN.2005.844857

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

We propose an innovative learning algorithm for enhancing the generalization ability of support vector machines (SVMs), when the Gausssian radial basis function (RBF) is used and when the parameter sigma, is very small. As learning patterns it uses not only the prescribed learning patterns but also newly inserted patterns in their neighbourhoods. In spite of the inserted many patterns, the size of the proposed optimization problem can be reduced to be same as the original one by using the averaging method. Many simulation results show the effectiveness of the proposed algorithm.

DOI： 10.1109/ISCAS.2005.1465416

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

We propose an innovative learning algorithm for enhancing the generalization ability of support vector machines (SVMs), when the Gausssian radial basis function (RBF) is used and when the parameter σ is very small. As learning patterns it uses not only the prescribed learning patterns but also newly inserted patterns in their neighbourhoods. In spite of the inserted many patterns, the size of the proposed optimization problem can be reduced to be same as the original one by using the averaging method. Many simulation results show the effectiveness of the proposed algorithm. © 2005 IEEE.

DOI： 10.1109/ISCAS.2005.1465416

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

A simple method of realizing a prescribed limit cycle by means of a continuous-time recurrent neural network with the piecewise-linear activation function is proposed. Since the proposed method is formulated as a set of linear inequalities, we can easily determine parameters of the network by using, for example, the simplex method.

DOI： 10.1109/ECCTD.2005.1523047

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）  

A new sufficient condition for three-cell cellular neural networks (CNNs) with the opposite-sign template to be completely stable is presented in this paper. It is also shown that even though the self-coupling is greater than unity the state trajectory which never enters a total saturation region can exist. © 2005 IEEE.

DOI： 10.1109/ISCAS.2005.1465675

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

A new sufficient condition for three-cell cellular neural networks (CNNs) with the opposite-sign template to be completely stable is presented in this paper. It is also shown that even though the self-coupling is greater than unity the state trajectory which never enters a total saturation region can exist.

DOI： 10.1109/ISCAS.2005.1465675

Web of Science

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

A novel method for training support vector machines (SVMs) is proposed to speed up the SVMs in test phase. It has three main steps. First, an SVM is trained on all the training samples, thereby producing a number of support vectors. Second, the support vectors, which contribute less to the shape of the decision surface, are excluded from the training set. Finally, the SVM is re-trained only on the remaining samples. Compared to the initially trained SVM, the efficiency of the finally trained SVM is highly improved, without system degradation.

DOI： 10.1109/ECCTD.2005.1523140

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

In this paper we give the necessary and sufficient conditions for 2-dimensional discrete-time systems described by the signum function to be stable.

Web of Science

researchmap

researchmap

researchmap

researchmap

DOI： 10.1007/978-3-540-28647-9_79

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

Convergence property of decomposition methods for solving quadratic optimization problems arising in support vector machines is studied. It is shown that under a mild condition any decomposition method converges to an optimal solution.

DOI： 10.1109/ISCAS.2004.1329911

Web of Science

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：IEEE  

This paper presents some properties of stable patterns that can be realized by a certain type of one-dimensional two-layer cellular neural networks (CNNs). We first introduce a notion of admissible local pattern (ALP) set. All the stable patterns of a CNN can be completely determined by the ALP set. We next show that all of 256 possible ALP sets can be realized by twolayer CNNs, while only 59 can be realized by single-layer CNNs. This means two-layer CNNs have a much higher potential for signal processing than single-layer CNNs.

DOI： 10.1109/MWSCAS.2004.1354008

Web of Science

researchmap

DOI： 10.1109/MWSCAS.2004.1354009

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

Realization of associative memories by cellular neural networks (CNNs) with binary output is studied. Concerning this problem, a CNN design method based upon generalized eigenvalue minimization (GEVM) has recently been proposed. In this brief, a new CNN design method which is based on the GEVM-based method will be presented. We first give some analytical results related to the basin of attraction of a memory vector. We then derive the design method by combining these analytical results and the GEVM-based method. We finally show through computer simulations that the proposed method can achieve higher recall probability than the original GEVM-based method.

DOI： 10.1109/TCSI.2003.819827

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

Realization of associative memories by cellular neural networks (CNNs) with binary output is studied. Concerning this problem, a CNN design method based upon generalized eigenvalue minimization (GEVM) has recently been proposed. In this brief, a new CNN design method which is based on the GEVM-based method will be presented. We first give some analytical results related to the basin of attraction of a memory vector. We then derive the design method by combining these analytical results and the GEVM-based method. We finally show through computer simulations that the proposed method can achieve higher recall probability than the original GEVM-based method.

DOI： 10.1109/TCSI.2003.819827

Scopus

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

Analog Hopfield neural networks (HNNs) have so far been used to solve many kinds of optimization problems, in particular, combinatorial problems such as the TSP, which call be described by an objective function and some equality constraints. When we solve a minimization problem with equality constraints by using HNNs, however, the constraints are satisfied only approximately. In this paper we propose a circuit which rigorously realizes the equality constraints and whose energy function corresponds to the prescribed objective function. We use the SPICE program to solve circuit equations corresponding to the above circuits. The proposed method is applied to several kinds of optimization problems and the results are very satisfactory.

Web of Science

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

This paper gives a new sufficient condition for cellular neural networks with delay (DCNNs) to be completely stable. The result is a generalization of two existing stability conditions for DCNNs, and also contains a complete stability condition for standard CNNs as a special case. Our new sufficient condition does not require the uniqueness of equilibrium point of DCNNs and is independent of the length of delay.

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

This paper investigates the behavior of one-dimensional discrete-time binary cellular neural networks with both the A- and B-templates and gives the necessary and sufficient conditions for the above network to be stable for unspecified fixed boundaries.

Web of Science

researchmap

DOI： 10.5687/iscie.15.600

researchmap

記述言語：英語   掲載種別：研究論文（国際会議プロシーディングス）   出版者・発行元：WORLD SCIENTIFIC PUBL CO PTE LTD  

This paper presents a design technique which is used to realize associative memories via cellular neural networks. The proposed method can store every prototype vector as a memory vector and maximize the areas of basin of attraction of memory vectors in a certain sense. The network parameters are obtained by solving optimization problems known as generalized eigenvalue problems. Simulation results proves that our method is better than the existing ones.

Web of Science

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）  

We give the necessary and sufficient conditions for a one-dimensional discrete-time binary cellular neural networks with both the A- and B-templates to be stable for a fixed boundary.

DOI： 10.1109/ISCAS.2002.1009920

Scopus

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

DOI： 10.1109/81.852931

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

This paper gives a new sufficient condition for complete stability of a nonsymmetric cellular neural network (CNN). The convergence theorem of the Gauss-Seidel method, which is an iterative technique for solving a linear algebraic equation, plays an important role in our proof. It is also shown that the existence of a stable equilibrium point does not imply complete stability of a nonsymmetric CNN.

DOI： 10.1109/81.703843

Web of Science

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：SCRIPTA TECHNICA-JOHN WILEY & SONS  

A rigorous method to determine whether a polynomial f(x) is nonnegative for all x in an interval is proposed, Conventionally, a method which determines whether the multiplicity of every zero of f(x) is even has been used for this problem. The conventional method is, however, somewhat complicated because the Sturm function sequence must be generated repeatedly and the number of variations of sign must be investigated for each sequence. The method proposed in this paper determines whether f(x) + epsilon (epsilon is a sufficiently small positive number) is positive for all x in the interval. This method is simpler than the conventional one in the sense that only two Sturm function sequences art. generated. Moreover, the determination can easily be made with the number of variations of sign of the function sequence made by concatenating two Sturm function sequences. (C) 1998 Scripta Technica.

Web of Science

researchmap

記述言語：英語   出版者・発行元：IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC  

This letter gives a new sufficient condition for nonsymmetric CNN's to have at least one stable equilibrium point. Existence of a stable equilibrium point is important for nonsymmetric CNN's because it is a necessary condition for complete stability. It is shown that our sufficient condition is a generalization of a previous result concerning the existence of a stable equilibrium point, and that it can easily be applied to space-invariant CNN's with a 3 x 3 neighborhood.

DOI： 10.1109/81.641777

Web of Science

researchmap

researchmap

researchmap

記述言語：英語   掲載種別：研究論文（学術雑誌）   出版者・発行元：IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG  

The number of solutions of a nonlinear equation x = sgn(Wx) is discussed. The equation is derived for the determination of equilibrium points of a kind of Hopfield neural networks. We impose some conditions on W. The conditions correspond to the case where a Hopfield neural network has n neurons arranged on a ring, each neuron has connections only from k preceding neurons and the magnitude of k connections decrease as the distance between two neurons increases. We show that the maximum number of solutions for the above case is extremely few and is independent of the number of neurons, n, if k is less than or equal to 4. We also show that the number of solutions generally increases exponentially with n by considering the case where k = n - 1.

Web of Science

researchmap

開催年月日： 2020年5月15日

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

開催年月日： 2020年5月15日

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

開催年月日： 2020年3月17日 - 2020年3月20日

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

開催年月日： 2020年3月17日 - 2020年3月20日

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

開催年月日： 2020年3月9日 - 2020年3月10日

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

開催年月日： 2020年3月9日 - 2020年3月10日

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（招待・特別）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：ポスター発表  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：ポスター発表  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：ポスター発表  

researchmap

記述言語：英語   会議種別：ポスター発表  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：ポスター発表  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：日本語   会議種別：口頭発表（一般）  

researchmap

記述言語：英語   会議種別：ポスター発表  

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap

researchmap