• Proceedings of the KIEE Conference
• /
• 1998.07b
• /
• pp.623-625
• /
• 1998

This paper focuses on the fast convergence in nonlinear parameter optimization which is necessary for the fitting of nonlinear models to data. The simulated annealing(SA) and genetic algorithm(GA), which are widely used for combinatorial optimization problems, are stochastic strategy for search of the ground state and a powerful tool for optimization. However, their main disadvantage is the long convergence time by unnecessary extra works. It is also recognised that gradient-based nonlinear programing techniques would typically fail to find global minimum. Therefore, this paper develops a modified SA which is the SDS(Stochastic deterministic stochastic) algorithm can minimize cost function of optimal problem.

• Management Science and Financial Engineering
• /
• v.21 no.1
• /
• pp.11-17
• /
• 2015

Min-max stochastic optimization is an approach to address the distribution ambiguity of the underlying random variable. We present a unified approach to the problem which utilizes the theory of convex order on the random variables. First, we consider a general framework for the problem and give a condition under which the convex order can be utilized to transform the min-max optimization problem into a simple minimization problem. Then extremal distributions are presented for some interesting classes of distributions. Finally, applications to the single-period inventory control problems are given.

• Journal of the Korea Society for Simulation
• /
• v.9 no.2
• /
• pp.89-102
• /
• 2000

In this paper, we consider an optimal allocation of constant service efforts in queueing network to maximize the system throughput. For this purpose, using the perturbation analysis, we apply a stochastic optimization algorithm to two types of queueing systems. Our simulation results indicate that the estimates obtained from a stochastic optimization algorithm for a two-tandem queuing network are very accurate, and those for closed loop manufacturing system are a little apart from the known optimal allocation. We find that as simulation time increases for obtaining a new gradient (performance measure with respect to decision variables) by perturbation algorithm, the estimates tend to be more stable. Thus, we consider that it would be more desirable to have more accurate sensitivity of performance measure by enlarging simulation time rather than more searching steps with less accurate sensitivity. We realize that more experiments on various types of systems are needed to identify such a relationship with regards to stopping rule, the size of moving step, and updating period for sensitivity.

• Advances in aircraft and spacecraft science
• /
• v.4 no.3
• /
• pp.297-316
• /
• 2017

With rapid growth in the complexity of large scale engineering systems, the application of multidisciplinary analysis and design optimization (MDO) in the engineering design process has garnered much attention. MDO addresses the challenge of integrating several different disciplines into the design process. Primary challenges of MDO include computational expense and poor scalability. The introduction of a distributed, collaborative computational environment results in better utilization of available computational resources, reducing the time to solution, and enhancing scalability. SORCER, a Java-based network-centric computing platform, enables analyses and design studies in a distributed collaborative computing environment. Two different optimization algorithms widely used in multidisciplinary engineering design-VTDIRECT95 and QNSTOP-are implemented on a SORCER grid. VTDIRECT95, a Fortran 95 implementation of D. R. Jones' algorithm DIRECT, is a highly parallelizable derivative-free deterministic global optimization algorithm. QNSTOP is a parallel quasi-Newton algorithm for stochastic optimization problems. The purpose of integrating VTDIRECT95 and QNSTOP into the SORCER framework is to provide load balancing among computational resources, resulting in a dynamically scalable process. Further, the federated computing paradigm implemented by SORCER manages distributed services in real time, thereby significantly speeding up the design process. Part 1 covers SORCER and the algorithms, Part 2 presents results for aircraft panel design with curvilinear stiffeners.

• Advances in aircraft and spacecraft science
• /
• v.4 no.3
• /
• pp.317-334
• /
• 2017

With rapid growth in the complexity of large scale engineering systems, the application of multidisciplinary analysis and design optimization (MDO) in the engineering design process has garnered much attention. MDO addresses the challenge of integrating several different disciplines into the design process. Primary challenges of MDO include computational expense and poor scalability. The introduction of a distributed, collaborative computational environment results in better utilization of available computational resources, reducing the time to solution, and enhancing scalability. SORCER, a Java-based network-centric computing platform, enables analyses and design studies in a distributed collaborative computing environment. Two different optimization algorithms widely used in multidisciplinary engineering design-VTDIRECT95 and QNSTOP-are implemented on a SORCER grid. VTDIRECT95, a Fortran 95 implementation of D. R. Jones' algorithm DIRECT, is a highly parallelizable derivative-free deterministic global optimization algorithm. QNSTOP is a parallel quasi-Newton algorithm for stochastic optimization problems. The purpose of integrating VTDIRECT95 and QNSTOP into the SORCER framework is to provide load balancing among computational resources, resulting in a dynamically scalable process. Further, the federated computing paradigm implemented by SORCER manages distributed services in real time, thereby significantly speeding up the design process. Part 1 covers SORCER and the algorithms, Part 2 presents results for aircraft panel design with curvilinear stiffeners.

• Geomechanics and Engineering
• /
• v.7 no.5
• /
• pp.525-537
• /
• 2014

The typical design of ground improvement with prefabricated vertical drains (PVD) and surcharge preloading involves a series of deterministic analyses using averaged or mean soil properties for the various combination of the PVD spacing and surcharge preloading height that would meet the criteria for minimum consolidation time and required degree of consolidation. The optimum design combination is then selected in which the total cost of ground improvement is a minimum. Considering the variability and uncertainties of the soil consolidation parameters, as well as considering the effects of soil disturbance (smear zone) and drain resistance in the analysis, this study presents a stochastic cost optimization of ground improvement with PVD and surcharge preloading. Direct Monte Carlo (MC) simulation and importance sampling (IS) technique is used in the stochastic analysis by limiting the sampled random soil parameters within the range from a minimum to maximum value while considering their statistical distribution. The method has been verified in a case study of PVD improved ground with preloading, in which average results of the stochastic analysis showed a good agreement with field monitoring data.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.4
• /
• pp.145-154
• /
• 1994

This paper proposes an efficient method for improving the training performance of the neural network by using a hybrid of a stochastic approximation and a backpropagation algorithm. The proposed method improves the performance of the training by appliying a global optimization method which is a hybrid of a stochastic approximation and a backpropagation algorithm. The approximate initial point for a stochastic approximation and a backpropagation algorihtm. The approximate initial point for fast global optimization is estimated first by applying the stochastic approximation, and then the backpropagation algorithm, which is the fast gradient descent method, is applied for a high speed global optimization. And further speed-up of training is made possible by adjusting the training parameters of each of the output and the hidden layer adaptively to the standard deviation of the neuron output of each layer. The proposed method has been applied to the parity checking and the pattern classification, and the simulation results show that the performance of the proposed method is superior to that of the backpropagation, the Baba's MROM, and the Sun's method with randomized initial point settings. The results of adaptive adjusting of the training parameters show that the proposed method further improves the convergence speed about 20% in training.

• Bulletin of the Korean Mathematical Society
• /
• v.55 no.4
• /
• 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.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.8 no.1
• /
• pp.41-59
• /
• 1983

The purpose of this paper is to present a method of multi-objective, stochastic optimization in water resources system which investigates the development of potential non-normal deterministic equivalents for subsequent use in multiobjective stochastic programming methods, Given probability statement involving a function of several random variables, it is often possible to obtain a deterministic equivalent of it that does not include any orginal random variables. A Stochastic trade-off technique-MSTOT is suggested to help a decision maker achieve satisfactory levels for several objective functions. This makes use of deterministic equivalents to handle random variables in the objective functions. The emphasis is in the development of non-normal deterministic equivalents for use in multiobjective stochastic techniques.

• Journal of the Korea Society for Simulation
• /
• v.3 no.2
• /
• pp.57-67
• /
• 1994

For many practical and industrial optimization problems where some or all of the system components are stochastic, the objective functions cannot be represented analytically. Therefore, modeling by computer simulation is one of the most effective means of studying such complex systems. In this paper, with discussion of simulation optimization techniques, a case study in machining process for application of simulation optimization is presented. Most of optimization techniques can be classified as single-or multiple-response techniques. The optimization of single-response category, these strategies are gradient based search methods, stochastic approximate method, response surface method, and heuristic search methods. In the multiple-response category, there are basically five distinct strategies for treating the responses and finding the optimum solution. These strategies are graphical method, direct search method, constrained optimization, unconstrained optimization, and goal programming methods. The choice of the procedure to employ in simulation optimization depends on the analyst and the problem to be solved.