• 대한수학회보
• /
• 제47권2호
• /
• pp.231-249
• /
• 2010

Asymptotic behavior of three-dimensional mixed, periodic and rotating magnetohydrodynamic system is investigated as the Rossby number goes to zero. The system presents the difficulty to be singular and mixed, that is hyperbolic in the vertical direction and parabolic in the horizontal one. The divergence free condition and the spectral properties of the penalization operator are crucial in the proofs. The main tools are the energy method, the Schochet's method and products laws in anisotropic Sobolev spaces.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제16권4호
• /
• pp.293-298
• /
• 2016

Document summarization is an important task in various areas where the goal is to select a few the most descriptive sentences from a given document as a succinct summary. Even without training data of human labeled summaries, there has been several interesting existing work in the literature that yields reasonable performance. In this paper, within the same unsupervised learning setup, we propose a more principled learning framework for the document summarization task. Specifically we formulate an optimization problem that expresses the requirements of both faithful preservation of the document contents and the summary length constraint. We circumvent the difficult integer programming originating from binary sentence selection via continuous relaxation and the low entropy penalization. We also suggest an efficient convex-concave optimization solver algorithm that guarantees to improve the original objective at every iteration. For several document datasets, we demonstrate that the proposed learning algorithm significantly outperforms the existing approaches.

• Steel and Composite Structures
• /
• 제27권1호
• /
• pp.27-33
• /
• 2018

In this study, a mixed-interpolated tensorial component 4 nodes method (MITC4) is treated as a numerical analysis model for topology optimization using multiple materials assigned within Reissner-Mindlin plates. Multi-material optimal topology and shape are produced as alternative plate retrofit designs to provide reasonable material assignments based on stress distributions. Element density distribution contours of mixing multiple material densities are linked to Solid Isotropic Material with Penalization (SIMP) as a design model. Mathematical formulation of multi-material topology optimization problem solving minimum compliance is an alternating active-phase algorithm with the Gauss-Seidel version as an optimization model of optimality criteria. Numerical examples illustrate the reliability and accuracy of the present design method for multi-material topology optimization with Reissner-Mindlin plates using MITC4 elements and steel materials.

• 한국전산유체공학회지
• /
• 제13권3호
• /
• pp.35-43
• /
• 2008

A methodology for the simulation of compressible high Reynolds number flow over rigid and moving bodies on a structured Cartesian grid is described in this paper. The approach is based on a modified version of the Brinkman Penalization method. To avoid oscillations in the vicinity of the body and to simulate shcok-containing flows, a Weighted Essentially Non-Oscillatory scheme is used to discretize the spatial flux derivatives. For high Reynolds number viscous flow, two turbulence models of the two-equation Menter's SST URANS model and a two-equation Detached Eddy Simulation are implemented. Some simple flow examples are given to assess the accuracy of the technique. Finally, a moving grid capability is demonstrated.

• Journal of Mechanical Science and Technology
• /
• 제17권10호
• /
• pp.1496-1506
• /
• 2003

The SIMP (solid isotropic material with penalization) approach is perhaps the most popular density variable relaxation method in topology optimization. This method has been very successful in many applications, but the optimization solution convergence can be improved when new variables, not the direct density variables, are used as the design variables. In this work, we newly propose S-shape functions mapping the original density variables nonlinearly to new design variables. The main role of S-shape function is to push intermediate densities to either lower or upper bounds. In particular, this method works well with nonlinear mathematical programming methods. A method of feasible directions is chosen as a nonlinear mathematical programming method in order to show the effects of the S-shape scaling function on the solution convergence.

• 대한수학회보
• /
• 제55권4호
• /
• pp.1303-1315
• /
• 2018

For principal component analysis (PCA) to efficiently analyze large scale matrices, it is crucial to find a few singular vectors in cheaper computational cost and under lower memory requirement. To compute those in a fast and robust way, we propose a new stochastic method. Especially, we adopt the stochastic variance reduced gradient (SVRG) method [11] to avoid asymptotically slow convergence in stochastic gradient descent methods. For that purpose, we reformulate the PCA problem as a unconstrained optimization problem using a quadratic penalty. In general, increasing the penalty parameter to infinity is needed for the equivalence of the two problems. However, in this case, exact penalization is guaranteed by applying the analysis in [24]. We establish the convergence rate of the proposed method to a stationary point and numerical experiments illustrate the validity and efficiency of the proposed method.

• Industrial Engineering and Management Systems
• /
• 제13권4호
• /
• pp.442-448
• /
• 2014

Portfolio optimization in the presence of estimation error can be stabilized by incorporating norm-constraints; this result was shown by DeMiguel et al. (A generalized approach to portfolio optimization: improving performance by constraining portfolio norms, Management Science, 5, 798-812, 2009), who reported empirical performance better than numerous competing approaches. We extend the idea of norm-constraints by introducing a powerful enhancement, grouped selection for portfolio optimization. Here, instead of merely penalizing norms of the assets being selected, we penalize groups, where within a group assets are treated alike, but across groups, the penalization may differ. The idea of groupwise selection is grounded in statistics, but to our knowledge, it is novel in the context of portfolio optimization. Novelty aside, the real benefits of groupwise selection are substantiated by experiments; our results show that groupwise asset selection leads to strategies with lower variance, higher Sharpe ratios, and even higher expected returns than the ordinary norm-constrained formulations.

• Structural Engineering and Mechanics
• /
• 제43권6호
• /
• pp.795-821
• /
• 2012

The size and topology of geometrically nonlinear dome structures are optimized thereby minimizing both its entire weight & joint (node) displacements and maximizing load-carrying capacity. Design constraints are implemented from provisions of American Petroleum Institute specification (API RP2A-LRFD). In accordance with the proposed design constraints, the member responses computed by use of arc-length technique as a nonlinear structural analysis method are checked at each load increment. Thus, a penalization process utilized for inclusion of unfeasible designations to genetic search is correspondingly neglected. In order to solve this complex design optimization problem with multiple objective functions, Non-dominated Sorting Genetic Algorithm II (NSGA II) approach is employed as a multi-objective optimization tool. Furthermore, the flexibility of proposed optimization is enhanced thereby integrating an automatic dome generating tool. Thus, it is possible to generate three distinct sphere-shaped dome configurations with varying topologies. It is demonstrated that the inclusion of brace (diagonal) members into the geometrical configuration of dome structure provides a weight-saving dome designation with higher load-carrying capacity. The proposed optimization approach is recommended for the design optimization of geometrically nonlinear dome structures.

• 대한수학회지
• /
• 제57권5호
• /
• pp.1079-1101
• /
• 2020

This work is concerned with a geometric inverse problem in fluid mechanics. The aim is to reconstruct an unknown obstacle immersed in a Newtonian and incompressible fluid flow from internal data. We assume that the fluid motion is governed by the Stokes-Brinkmann equations in the two dimensional case. We propose a simple and efficient reconstruction method based on the topological sensitivity concept. The geometric inverse problem is reformulated as a topology optimization one minimizing a least-square functional. The existence and stability of the optimization problem solution are discussed. A topological sensitivity analysis is derived with the help of a straightforward approach based on a penalization technique without using the classical truncation method. The theoretical results are exploited for building a non-iterative reconstruction algorithm. The unknown obstacle is reconstructed using a levelset curve of the topological gradient. The accuracy and the robustness of the proposed method are justified by some numerical examples.

• 한국전산유체공학회:학술대회논문집
• /
• 한국전산유체공학회 2007년도 추계 학술대회논문집
• /
• pp.200-208
• /
• 2007

A methodology for the simulation of compressible high Reynolds number flow over rigid and moving bodies on a structured Cartesian grid is described in this paper. The approach is based on a modified version of the Brinkman Penalization method. To avoid oscillations in the vicinity of the body and to simulate shcok-containing flows, a Weighted Essentially Non-Oscillatory scheme is used to discretize the spatial flux derivatives. For high Reynolds number viscous flow, two turbulence models of the two-equation Menter's SST URANS model and a two-equation Detached Eddy Simulation are implemented. Some simple flow examples are given to assess the accuracy of the technique. Finally, a moving grid capability is demonstrated.