The Experts below are selected from a list of 3273 Experts worldwide ranked by ideXlab platform
Ann B Ragin - One of the best experts on this subject based on the ideXlab platform.
-
multi graph clustering based on Interior Node topology with applications to brain networks
European conference on Machine Learning, 2016Co-Authors: Bokai Cao, Jiawei Zhang, Ann B RaginAbstract:Learning from graph data has been attracting much attention recently due to its importance in many scientific applications, where objects are represented as graphs. In this paper, we study the problem of multi-graph clustering i.e., clustering multiple graphs. We propose a multi-graph clustering approach MGCT based on the Interior-Node topology of graphs. Specifically, we extract the Interior-Node topological structure of each graph through a sparsity-inducing Interior-Node clustering. We merge the Interior-Node clustering stage and the multi-graph clustering stage into a unified iterative framework, where the multi-graph clustering will influence the Interior-Node clustering and the updated Interior-Node clustering results will be further exerted on multi-graph clustering. We apply MGCT on two real brain network data sets i.e., ADHD and HIV. Experimental results demonstrate the superior performance of the proposed model on multi-graph clustering.
Bokai Cao - One of the best experts on this subject based on the ideXlab platform.
-
multi graph clustering based on Interior Node topology with applications to brain networks
European conference on Machine Learning, 2016Co-Authors: Bokai Cao, Jiawei Zhang, Ann B RaginAbstract:Learning from graph data has been attracting much attention recently due to its importance in many scientific applications, where objects are represented as graphs. In this paper, we study the problem of multi-graph clustering i.e., clustering multiple graphs. We propose a multi-graph clustering approach MGCT based on the Interior-Node topology of graphs. Specifically, we extract the Interior-Node topological structure of each graph through a sparsity-inducing Interior-Node clustering. We merge the Interior-Node clustering stage and the multi-graph clustering stage into a unified iterative framework, where the multi-graph clustering will influence the Interior-Node clustering and the updated Interior-Node clustering results will be further exerted on multi-graph clustering. We apply MGCT on two real brain network data sets i.e., ADHD and HIV. Experimental results demonstrate the superior performance of the proposed model on multi-graph clustering.
Jiawei Zhang - One of the best experts on this subject based on the ideXlab platform.
-
multi graph clustering based on Interior Node topology with applications to brain networks
European conference on Machine Learning, 2016Co-Authors: Bokai Cao, Jiawei Zhang, Ann B RaginAbstract:Learning from graph data has been attracting much attention recently due to its importance in many scientific applications, where objects are represented as graphs. In this paper, we study the problem of multi-graph clustering i.e., clustering multiple graphs. We propose a multi-graph clustering approach MGCT based on the Interior-Node topology of graphs. Specifically, we extract the Interior-Node topological structure of each graph through a sparsity-inducing Interior-Node clustering. We merge the Interior-Node clustering stage and the multi-graph clustering stage into a unified iterative framework, where the multi-graph clustering will influence the Interior-Node clustering and the updated Interior-Node clustering results will be further exerted on multi-graph clustering. We apply MGCT on two real brain network data sets i.e., ADHD and HIV. Experimental results demonstrate the superior performance of the proposed model on multi-graph clustering.
Robert Tibshirani - One of the best experts on this subject based on the ideXlab platform.
-
hierarchical clustering with prototypes via minimax linkage
Journal of the American Statistical Association, 2011Co-Authors: Jacob Bien, Robert TibshiraniAbstract:Agglomerative hierarchical clustering is a popular class of methods for understanding the structure of a dataset. The nature of the clustering depends on the choice of linkage—that is, on how one measures the distance between clusters. In this article we investigate minimax linkage, a recently introduced but little-studied linkage. Minimax linkage is unique in naturally associating a prototype chosen from the original dataset with every Interior Node of the dendrogram. These prototypes can be used to greatly enhance the interpretability of a hierarchical clustering. Furthermore, we prove that minimax linkage has a number of desirable theoretical properties; for example, minimax-linkage dendrograms cannot have inversions (unlike centroid linkage) and is robust against certain perturbations of a dataset. We provide an efficient implementation and illustrate minimax linkage’s strengths as a data analysis and visualization tool on a study of words from encyclopedia articles and on a dataset of images of human...
K Nakahashi - One of the best experts on this subject based on the ideXlab platform.
-
unstructured dynamic mesh for large movement and deformation
40th AIAA Aerospace Sciences Meeting & Exhibit, 2002Co-Authors: K Matsushima, M Murayama, K NakahashiAbstract:In this paper, a simple robust method of the unstructured dynamic mesh for three-dimensional moving and deforming body problems is proposed. The method is developed to avoid the generation of squashed invalid elements by considering each elemental shape. With several three-dimensional applications, it is demonstrated that the present method significantly improves the robustness for problems with large motions of bodies without much penalty in CPU time. The use of the present method for grid movements on curved surface is also discussed. Its capability is demonstrated for the surface grid movement in the tail-fuselage juncture region for the evaluation of the airplane control surfaces. INTRODUCTION In the last 30 years, the computational fluid dynamics (CFD) has achieved a significant progress and the CFD is considered to be very close to its maturing stage in the computation of flows around airplanes at designed conditions. However, the CFD is still far from being a maturing subject for geometrically complex and largely moving and deforming body problems. Such flows with body movement and deformation are often encountered at various important engineering problems such as flutter problems, rocket separation, parachute and balloon dynamics, and oscillating and flapping wings. Numerical simulation of flights of a maneuvering airplane including the aircraft response to control surface deflection and aeroelasticity is also important. To treat these problems, the grid must be moved accompanied with the body movement and deformation. For the simulation of flows with * Graduate student, Department of aeronautics and space engineering, Tohoku University f Professor, Department of aeronautics and space engineering, Tohoku University, Associate Fellow AIAA 1 Assistant professor, Department of aeronautics and space engineering, AIAA Member Copyright © 2002 by American Institute of Aeronautics and Astronautics, Inc. All rights reserved. moving rigid bodies, overset unstructured grid techniques [1,2], which use several unstructured grids to cover the flow field with the moving bodies, are very useful. On the other hand, for the simulation of flows with deforming bodies, the dynamic mesh method is required. Already several methods for the dynamic mesh have been developed and showed good results for the problems with small amplitude oscillation and deformation [3-8]. However, the methods may lack the robustness necessary for large deformation problems. Therefore, the dynamic mesh is commonly used with grid remeshing method [9] if the mesh has some distorted and invalid elements. By the use of the remeshing, the obtained grid may be clean. However, the remeshing has serious problems. It suffers from the loss of physical conservation law. In other words, it may locally reduce the computational accuracy due to drastic grid size variation. It also needs extra computational costs especially when solving a flow field about three-dimensional complex bodies. In this paper, a robust method for the dynamic mesh strategy is developed so as to minimize the requirement of the remeshing due to ill-conditioned cells. Most common approach of the dynamic meshes is to use the linear tension spring analogy. However, this approach cannot control the cell shape and often creates ill-conditioned cells for large movements. For structured grid, one of the present authors proposed a tension-torsion spring analogy for solution adaptive grid [10]. The torsion spring was very effective to keep the grid quality. Similar concept was employed for two-dimensional unstructured grid by Farhat et al. [11] in a sophisticated manner. Additional torsion springs were designed to prohibit the interpenetration of neighboring triangles by the consideration of the internal angles of triangles and showed good performance in terms of the robustness and quality and enlarged its range of applications for two-dimensional problems. However, the reliable and simple dynamic mesh method for three-dimensional problems has not been established yet. Another requirement of the dynamic mesh method is the treatment of the grid of the wing-fuselage juncture, when it is used in the design optimization 1 American Institute of Aeronautics and Astronautics (c)2002 American Institute of Aeronautics & Astronautics or Published with Permission of Author(s) and/or Author(s)' Sponsoring Organization. process of the wing-fuselage configuration and the computation of the maneuver of aircraft in response to the moving control surface deflection. Because of the location and geometry change of a wing on a fuselage surface, the topological structure of the surface grid of the wing-fuselage juncture should be changed. To treat it, surface grid points may have to be moved. But it is difficult to directly move the surface grid points along the curved surface properly. In this paper, a simple method of the unstructured dynamic mesh for three-dimensional moving and deforming body problems is proposed. The robustness of the method is evaluated for several three-dimensional problems with largely moving and deforming bodies. In addition, a method to apply the dynamic mesh to the surface mesh movement is discussed. Such a mesh movement on the curved surface is required for a simulation of airplane response to the movement of the control surfaces. MESH POINT MOVEMENT STRATEGIES Although there are several methods to move the mesh points, linear tension spring analogy method [3,4] is typically used to move the mesh points because of its simplicity and low computational costs. In the method, each edge of the grid is modeled as a linear tension spring. The forces F^^-are generated to these springs at each vertex by the mesh point movement, * spring ij ~ *spring ij (1) where kspringijis a spring stiffness coefficient at edge ij and Ax// is a displacement. The static equilibrium equations, Eq. (2), of these forces are solved iteratively at each Interior Node and the displacements in x, y, and z directions are determined. If: spring ij,=0 (2) Here, grid points on the outer boundary of the mesh are fixed and the displacement on the body surface is given as the initial conditions. By the use of the edge length as the spring stiffness described at Eq. (3), the method can prevent two vertices from colliding.