• Journal of applied mathematics & informatics
• /
• v.14 no.1_2
• /
• pp.123-136
• /
• 2004

In this paper, a combining trust region and line search algorithm for equality constrained optimization is proposed. At each iteration, we only need to solve the trust region subproblem once, when the trust region trial step can not be accepted, we switch to line search to obtain the next iteration. Hence, the difficulty of repeated solving trust region subproblem in an iterate is avoided. In order to allow the direction of negative curvature, we add second correction step in trust region step and employ nonmonotone technique in line search. The global convergence and local superlinearly rate are established under certain assumptions. Some numerical examples are given to illustrate the efficiency of the proposed algorithm.

• Journal of applied mathematics & informatics
• /
• v.27 no.1_2
• /
• pp.233-243
• /
• 2009

This paper concerns an algorithm that combines line search and trust region step for nonlinear optimization problems. Unlike traditional trust region methods, we incorporate the Armijo line search technique into trust region method to solve the subproblem. In addition, the subproblem is solved accurately, but instead solved by inaccurate method. If a trial step is not accepted, our algorithm performs the Armijo line search from the failed point to find a suitable steplength. At each iteration, the subproblem is solved only one time. In contrast to interior methods, the optimal solution is derived by iterating from outside of the feasible region. In numerical experiment, we apply the algorithm to nonlinear portfolio optimization problems, primary numerical results are presented.

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.04b
• /
• pp.721-723
• /
• 2004

A trust-region method is a quite attractive optimization technique. It is, in general, faster than the steepest descent method and is free of a learning rate unlike the gradient-based methods. In addition to its convergence property (between linear and quadratic convergence), ifs stability is always guaranteed, in contrast to the Newton's method. In this paper, we present an efficient implementation of the maximum likelihood independent component analysis (ICA) using the trust-region method, which leads to trust-region-based ICA (TR-ICA) algorithms. The useful behavior of our TR-ICA algorithms is confimed through numerical experimental results.

• Journal of applied mathematics & informatics
• /
• v.27 no.5_6
• /
• pp.981-1000
• /
• 2009

This paper considers asset liability management problems when their deterministic equivalent formulations are general nonlinear optimization problems. The presented approach uses a nonmonotone trust region strategy for solving a sequence of unconstrained subproblems parameterized by a scalar parameter. The objective function of each unconstrained subproblem is an augmented penalty-barrier function that involves both primal and dual variables. Each subproblem is solved approximately. The algorithm does not restrict a monotonic decrease of the objective function value at each iteration. If a trial step is not accepted, the algorithm performs a non monotone line search to find a new acceptable point instead of resolving the subproblem. We prove that the algorithm globally converges to a point satisfying the second-order necessary optimality conditions.

• Journal of the Korean Mathematical Society
• /
• v.54 no.2
• /
• pp.613-626
• /
• 2017

For the overdetermined linear system, when both the data matrix and the observed data are contaminated by noise, Total Least Squares method is an appropriate approach. Since an ill-conditioned data matrix with noise causes a large perturbation in the solution, some kind of regularization technique is required to filter out such noise. In this paper, we consider a Dual regularized Total Least Squares problem. Unlike the Tikhonov regularization which constrains the size of the solution, a Dual regularized Total Least Squares problem considers two constraints; one constrains the size of the error in the data matrix, the other constrains the size of the error in the observed data. Our method derives two nonlinear equations to construct the iterative method. However, since the Jacobian matrix of two nonlinear equations is not guaranteed to be nonsingular, we adopt a trust-region based iteration method to obtain the solution.

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.10b
• /
• pp.697-699
• /
• 2004

Stochastic Neighbor Embedding(SNE) is a probabilistic method of mapping high-dimensional data space into a low-dimensional representation with preserving neighbor identities. Even though SNE shows several useful properties, the gradient-based naive SNE algorithm has a critical limitation that it is very slow to converge. To overcome this limitation, faster optimization methods should be considered by using trust region method we call this method fast TR SNE. Moreover, this paper presents a couple of useful optimization methods(i.e. conjugate gradient method and Newton's method) to embody fast SNE algorithm. We compared above three methods and conclude that TR-SNE is the best algorithm among them considering speed and stability. Finally, we show several visualizing experiments of TR-SNE to confirm its stability by experiments.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.29 no.6 s.237
• /
• pp.842-851
• /
• 2005

In this paper, a new method for constrained optimization of noisy functions is proposed. In approximate optimization using response surface methods, if constraints have severe noise, the approximate feasible region defined by approximate constraints is apt to include some of the infeasible region defined by actual constraints. This can cause the approximate optimum to converge into the infeasible region. In the proposed method, the approximate optimization is performed with the approximate constraints shifted by their deviations, which are calculated using a diagonal quadratic response surface method. This can prevent the approximate optimum from converging into the infeasible region. To fit the objective and constraints into diagonal quadratic models, we select the center and 4 additional points along each axis of design variables as experimental points. The deviation of each function is calculated using the differences between the real and approximate function values at the experimental points. A sequential approximate optimization technique based on the trust region algorithm is adopted to manage approximate models. The proposed approach is validated by solving some design problems. The results of the problems show the effectiveness of the proposed method.

• Industrial Engineering and Management Systems
• /
• v.9 no.1
• /
• pp.10-19
• /
• 2010

This paper presents an extended computing procedure for the global optimization of the triple response system (TRS) where the response functions are nonconvex (nonconcave) quadratics and the input factors satisfy a radial region of interest. The TRS arising from response surface modeling can be approximated using a nonlinear mathematical program involving one primary (objective) function and two secondary (constraints) functions. An optimization algorithm named triple response surface algorithm (TRSALG) is proposed to determine the global optimum for the nondegenerate TRS. In TRSALG, the Lagrange multipliers of target (secondary) functions are computed by using the Hooke-Jeeves search method, and the Lagrange multiplier of the radial constraint is located by using the trust region (TR) method at the same time. To ensure global optimality that can be attained by TRSALG, included is the means for detecting the degenerate case. In the field of numerical optimization, as the family of TR approach always exhibits excellent mathematical properties during optimization steps, thus the proposed algorithm can guarantee the global optimal solution where the optimality conditions are satisfied for the nondegenerate TRS. The computing procedure is illustrated in terms of examples found in the quality literature where the comparison results with a gradient-based method are used to calibrate TRSALG.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.29 no.9 s.240
• /
• pp.1199-1208
• /
• 2005

Nowadays, it is performed actively to optimize by using an approximate model. This is called the approximate optimization. In addition, the sequential approximate optimization (SAO) is the repetitive method to find an optimum by considering the convergence of an approximate optimum. In some recent studies, it is proposed to increase the fidelity of approximate models by applying the sequential sampling. However, because the accuracy and efficiency of an approximate model is directly connected with the design area and the termination criteria are not clear, sequential sampling method has the disadvantages that could support an unreasonable approximate optimum. In this study, the SAO is executed by using trust region, Kriging model and Optimal Latin Hypercube design (OLHD). Trust region is used to guarantee the convergence and Kriging model and OLHD are suitable for computer experiment. finally, this SAO method is applied to various optimization problems of highly nonlinear mathematical functions. As a result, each approximate optimum is acquired and the accuracy and efficiency of this method is verified by comparing with the result by established method.

• Journal of the Korean Society for Precision Engineering
• /
• v.30 no.10
• /
• pp.1031-1039
• /
• 2013

This paper presents gain optimization of a 6-DOF underwater robotic platform with 4 rotatable thrusters. To stabilize the 6-DOF motion of the underwater robotic platform, a back-stepping controller is designed with 6 proportional gains and 6 derivative gains. The 12 gains of the backstepping controller are optimized to decrease settling time in step response in 6-DOF motion independently. Stability criterion and overshoots are used as a constraint of the optimization problem. Trust-region algorithm and hybrid Taguchi-Random order Coordinate search algorithm are used to determine the optimal parameters, and the results by two methods are analyzed. Additionally, the resulting controller shows improved performance under disturbances.