• Journal of the Korean Statistical Society
• /
• v.20 no.2
• /
• pp.99-107
• /
• 1991

Behrens-Fisher problem is viewed from a model selection approach. Normal distribution is regarded as an approximating model, A criterion, called TIC, is derived and is compared with selection criteria such as AIC and a bootstrap estimator. Stochastic approximation is used since no closed form expression is available for the bootstrap estimator.

• Bulletin of the Korean Mathematical Society
• /
• v.44 no.1
• /
• pp.125-130
• /
• 2007

We propose a simple and intuitive method to derive the exact convergence rate of global $L_{2}-norm$ error for strong numerical approximation of stochastic differential equations the result of which has been reported by Hofmann and $M{\"u}ller-Gronbach\;(2004)$. We conclude that any strong numerical scheme of order ${\gamma}\;>\;1/2$ has the same optimal convergence rate for this error. The method clearly reveals the structure of global $L_{2}-norm$ error and is similarly applicable for evaluating the convergence rate of global uniform approximations.

• Transactions on Control, Automation and Systems Engineering
• /
• v.4 no.2
• /
• pp.124-129
• /
• 2002

This paper demonstrates that the largest Lyapunov exponent λ of recurrent neural networks can be controlled efficiently by a stochastic gradient method. An essential core of the proposed method is a novel stochastic approximate formulation of the Lyapunov exponent λ as a function of the network parameters such as connection weights and thresholds of neural activation functions. By a gradient method, a direct calculation to minimize a square error (λ - λ$\^$obj/)$^2$, where λ$\^$obj/ is a desired exponent value, needs gradients collection through time which are given by a recursive calculation from past to present values. The collection is computationally expensive and causes unstable control of the exponent for networks with chaotic dynamics because of chaotic instability. The stochastic formulation derived in this paper gives us an approximation of the gradients collection in a fashion without the recursive calculation. This approximation can realize not only a faster calculation of the gradient, but also stable control for chaotic dynamics. Due to the non-recursive calculation. without respect to the time evolutions, the running times of this approximation grow only about as N$^2$ compared to as N$\^$5/T that is of the direct calculation method. It is also shown by simulation studies that the approximation is a robust formulation for the network size and that proposed method can control the chaos dynamics in recurrent neural networks efficiently.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.5
• /
• pp.98-106
• /
• 1998

This paper proposes an efficient learning algorithm for improving the training performance of the neural network. The proposed method improves the training performance by applying the backpropagation algorithm of a global optimization method which is a hybrid of a stochastic approximation and a conjugate gradient method. The approximate initial point for f a ~gtl obal optimization is estimated first by applying the stochastic approximation, and then the conjugate gradient method, which is the fast gradient descent method, is applied for a high speed optimization. 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 those of the conventional backpropagation and the backpropagation algorithm which is a hyhrid of the stochastic approximation and steepest descent method.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.602-606
• /
• 2005

A systematic procedure of probability analysis for general distributions is developed based on the first four moments estimated from polynomial interpolation of the system response function and the Pearson system. The function approximation is based on a specially selected experimental region for accuracy and the number of function evaluations is taken equal to that of the unknown coefficient for efficiency. For this purpose, three error-minimizing conditions are proposed and corresponding canonical experimental regions are formed for popular probability. This approach is applied to study the stochastic properties of the performance functions of a MEMS structure, which has quite large fabrication errors compared to other structures. Especially, the vibratory micro-gyroscope is studied using the statistical moments and probability density function (PDF) of the performance function to be the difference between resonant frequencies corresponding to sensing and driving mode. The results show that it is very sensitive to the fabrication errors and that the types of PDF of each variable also affect the stochastic properties of the performance function although they have same the mean and variance.

• International Journal of Control, Automation, and Systems
• /
• v.6 no.4
• /
• pp.559-570
• /
• 2008

This paper proposes an improved particle swarm optimization named PSO with Controllable Random Exploration Velocity (PSO-CREV) behaving an additional exploration behavior. Different from other improvements on PSO, the updating principle of PSO-CREV is constructed in terms of stochastic approximation diagram. Hence a stochastic velocity independent on cognitive and social components of PSO can be added to the updating principle, so that particles have strong exploration ability than those of conventional PSO. The conditions and main behaviors of PSO-CREV are described. Two properties in terms of "divergence before convergence" and "controllable exploration behavior" are presented, which promote the performance of PSO-CREV. An experimental method based on a complex test function is proposed by which the proper parameters of PSO-CREV used in practice are figured out, which guarantees the high exploration ability, as well as the convergence rate is concerned. The benchmarks and applications on FCRNN training verify the improvements brought by PSO-CREV.

• Journal of Institute of Control, Robotics and Systems
• /
• v.6 no.3
• /
• pp.263-272
• /
• 2000

In this paper we give some geometric condition for a stochastic nonlinear system and we propose a control method for a stochastic nonlinear system using neural networks. Since a competitive learning neural networks has been developed based on the stochastcic approximation method it is regarded as a stochastic recursive filter algorithm. In addition we provide a filtering and control condition for a stochastic nonlinear system called the perfect filtering condition in a viewpoint of stochastic geometry. The stochastic nonlinear system satisfying the perfect filtering condition is decoupled with a deterministic part and purely semi martingale part. Hence the above system can be controlled by conventional control laws and various intelligent control laws. Computer simulation shows that the stochastic nonlinear system satisfying the perfect filtering condition is controllable and the proposed neural controller is more efficient than the conventional LQG controller and the canonical LQ-Neural controller.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.346-352
• /
• 1998

In this paper, we give some geometric condition for a stochastic nonlinear system and we propose a control method for a stochastic nonlinear system using neural networks. Since a competitive learning neural networks has been developed based on the stochastic approximation method, it is regarded as a stochastic recursive filter algorithm. In addition, we provide a filtering and control condition for a stochastic nonlinear system, called perfect filtering condition, in a viewpoint of stochastic geometry. The stochastic nonlinear system satisfying the perfect filtering condition is decoupled with a deterministic part and purely semi martingale part. Hence, the above system can be controlled by conventional control laws and various intelligent control laws. Computer simulation shows that the stochastic nonlinear system satisfying the perfect filtering condition is controllable. and the proposed neural controller is more efficient than the conventional LQG controller and the canoni al LQ-Neural controller.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2008.10a
• /
• pp.65-75
• /
• 2008

Over the last three decades, there are many researches focusing on the practice and theory of RM in airlines. Most of them have dealt with a seat assignment problem for maximizing the total revenue. In this study, we focus on a seat assignment problem in airlines. The seat assignment problem can be modeled as a stochastic programming model which is difficulty to solve optimally. However, with some assumptions on the demand distribution functions and a linear approximation technique, we can transform the complex stochastic programming model to a Linear Programming model. Some computational experiments are performed to evaluate out model with randomly generated data. They show that our model has a good performance comparing to existing models, and can be considered as a basis for further studies on improving existing seat assignment models.

• Korean Management Science Review
• /
• v.26 no.3
• /
• pp.117-131
• /
• 2009

Over the last three decades, there are many researches focusing on the practice and theory of RM in airlines. Most of them have dealt with a seat assignment problem for maximizing the total revenue. In this study, we focus on a seat assignment problem in airlines. The seat assignment problem can be modeled as a stochastic programming model which is difficulty to solve optimally. However, with some assumptions on the demand distribution functions and a linear approximation technique, we can transform the complex stochastic programming model to a Linear Programming model. Some computational experiments are performed to evaluate out model with randomly generated data. They show that our model has a good performance comparing to existing models, and can be considered as a basis for further studies on improving existing seat assignment models.