• Communications for Statistical Applications and Methods
• /
• v.27 no.1
• /
• pp.129-140
• /
• 2020

In this paper, we introduce an efficient algorithm for the non-convex penalized multinomial logistic regression that can be uniformly applied to a class of non-convex penalties. The class includes most non-convex penalties such as the smoothly clipped absolute deviation, minimax concave and bridge penalties. The algorithm is developed based on the concave-convex procedure and modified local quadratic approximation algorithm. However, usual quadratic approximation may slow down computational speed since the dimension of the Hessian matrix depends on the number of categories of the output variable. For this issue, we use a uniform bound of the Hessian matrix in the quadratic approximation. The algorithm is available from the R package ncpen developed by the authors. Numerical studies via simulations and real data sets are provided for illustration.

• Communications for Statistical Applications and Methods
• /
• v.19 no.5
• /
• pp.655-662
• /
• 2012

Classification is an important research field in pattern recognition with high-dimensional predictors. The support vector machine(SVM) is a penalized feature selector and classifier. It is based on the hinge loss function, the non-convex penalty function, and the smoothly clipped absolute deviation(SCAD) suggested by Fan and Li (2001). We developed the algorithm for the multiclass SVM with the SCAD penalty function using the local quadratic approximation. For multiclass problems we compared the performance of the SVM with the $L_1$, $L_2$ penalty functions and the developed method.

• Communications for Statistical Applications and Methods
• /
• v.20 no.6
• /
• pp.481-490
• /
• 2013

Classification is an important research area as data can be easily obtained even if the number of predictors becomes huge. The support vector machine(SVM) is widely used to classify a subject into a predetermined group because it gives sound theoretical background and better performance than other methods in many applications. The SVM can be viewed as a penalized method with the hinge loss function and penalty functions. Instead of $L_2$ penalty function Fan and Li (2001) proposed the smoothly clipped absolute deviation(SCAD) satisfying good statistical properties. Despite the ability of SVMs, they have drawbacks of non-robustness when there are outliers in the data. We develop a robust SVM method using a weight function with the SCAD penalty function based on the local quadratic approximation. We compare the performance of the proposed SVM with the SVM using the $L_1$ and $L_2$ penalty functions.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 1999.10a
• /
• pp.419-424
• /
• 1999

In this paper, the comparison of the first order approximation schemes such as SLP (sequential linear programming), CONLIN(convex linearization), MMA(method of moving asymptotes) and the second order approximation scheme, SQP(sequential quadratic programming) was accomplished for optimization of and nonlinear structures. It was found that MMA and SQP(sequential quadratic programming) was accomplished for optimization of and nonlinear structures. It was found that MMA and SQP are the most efficient methods for optimization. But the number of function call of SQP is much more than that of MMA. Therefore, when it is considered with the expense of computation, MMA is more efficient than SQP. In order to examine the efficiency of MMA for complex optimization problem, it was applied to the helicopter tail boom considering column buckling and local wall buckling constraints. It is concluded that MMA can be a very efficient approximation scheme from simple problems to complex problems.

• Communications for Statistical Applications and Methods
• /
• v.19 no.1
• /
• pp.135-145
• /
• 2012

It is well known that the maximum likelihood estimator(MLE) in normal mixture models with unequal variances does not fall in the interior of the parameter space. Recently, a doubly smoothed maximum likelihood estimator(DS-MLE) (Seo and Lindsay, 2010) was proposed as a general alternative to the ordinary maximum likelihood estimator. Although this method gives a natural modification to the ordinary MLE, its computation is cumbersome due to intractable integrations. In this paper, we derive an EM algorithm for the DS-MLE under normal mixture models and propose a fast computational tool using a local quadratic approximation. The accuracy and speed of the proposed method is then presented via some numerical studies.

• KIEE International Transaction on Systems and Control
• /
• v.12D no.1
• /
• pp.33-38
• /
• 2002

In this paper, the infinite time optimal regulation problem for weakly coupled bilinear systems with quadratic performance criteria is obtained by a sequence of algebraic Lyapunov equations. This is the new approach is based on the successive approximation. In particular, the order reduction is achieved by using suitable state transformation so that the original Lyapunov equations are decomposed into the reduced-order local Lyapunov equations. The proposed algorithms not only solve optimal control problems in the weakly coupled bilinear system but also reduce the computation time. This paper also includes an example to demonstrate the procedures.

• Journal of the Korean Society for Precision Engineering
• /
• v.23 no.8 s.185
• /
• pp.89-99
• /
• 2006

In this paper a new shape reconstruction method that allows us to construct surface models from very large sets of points is presented. In this method the global domain of interest is divided into smaller domains where the problem can be solved locally. These local solutions of subdivided domains are blended together according to weighting coefficients to obtain a global solution using partition of unity function. The suggested approach gives us considerable flexibility in the choice of local shape functions which depend on the local shape complexity and desired accuracy. At each domain, a quadratic polynomial function is created that fits the points in the domain. If the approximation is not accurate enough, other higher order functions including cubic polynomial function and RBF(Radial Basis Function) are used. This adaptive selection of local shape functions offers robust and efficient solution to a great variety of shape reconstruction problems.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 1995.09a
• /
• pp.211-219
• /
• 1995

블록-삼각(Block-angular)구조를 갖는 선형 제약식과 분리되지 않는 콘벡스 목적함수의 대규모 비분리 콘벡스 최적화 문제의 해법으로 번들-분해법 (Bundle Based Decomposition)을 이용한 알고리즘 SQA(Separable Quadratic Approximation)은 비분리 콘벡스 프로그램을 분리가능한 2차계획 법(Separable Quadratic Programming) 문제로 근사화시켜 번들-분해법을 축 차적으로 적용한다. 본 연구는 수렴성(local convergence & global convergence) 및 알고리즘 구현 [1]에 이어 이에 대한 수치적용 결과를 중심 으로 소개한다. 수치 적용은 ANSI C로 작성된 SQA 프로그램을 SUN SPARC II에서 실행하였으며 이때 대규모 비분리 최적화 문제의 비분리 목 적함수와 블록-삼각 구조의 선형 제약식들이 계수들은 ANSI C의 랜덤함수 로부터 임의의 값들을 이용하였다. 이와같은 다양한 비분리 콘벡스 최적화 문제에 대한 수렴성, 반복회수 및 처리시간등의 결과와 함께 GAMS/MINOS 의 최적해를 소개한다.

• Journal of the Korean Society of Manufacturing Technology Engineers
• /
• v.9 no.2
• /
• pp.59-65
• /
• 2000

In this paper the comparison of the first order approximation schemes such as SLP(sequential linear programming) CONLIN(convex linearization) MMA(method of moving asymptotes) and the second order approximation scheme SQP(sequential quadratic programming) was accomplished for optimization of nonlinear structures. It was found that MMA and SQP are the most efficient methods for optimization. But the number of function call of SQP is much more than that of MMA. Therefore when it is considered with the expense of computation MMA is more efficient than SQP. In order to examine the efficiency of MMA for complex optimization problem it was applied to the helicopter tail boom con-sidering column buckling and local wall buckling constraints. it is concluded that MMA can be a very efficient approxima-tion scheme from simple problems to complex problems.

• Proceedings of the KIEE Conference
• /
• 2001.07d
• /
• pp.1999-2001
• /
• 2001

The infinite time optimum to regulate the problem of weakly coupled bilinear systems with a quadratic performance criterion is obtained by a sequence of algebraic Lyapunov equations. The new approach is based on the successive approximations. In particular, the order reduction is achieved by using suitable state transformation so that the original Lyapunov equations are decomposed into the reduced-order local Lyapunov equations. The proposed algorithms not only solve optimal control problems in the weakly coupled bilinear system but also reduce the computation time. This paper also includes an example to demonstrate the procedures.