• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1997.10a
• /
• pp.301-308
• /
• 1997

The learning speed of the neural networks, the most important factor in applying to real problems, greatly depends on the learning rate of the networks, Three approaches-empirical, deterministic, and stochastic ones-have been proposed to date. We proposed a new learning-rate selection algorithm using an evolutionary programming search scheme. Even though the performance of our method showed better than those of the other methods, it was found that taking much time for selecting evolutionary learning rates made the performance of our method degrade. This was caused by using static intervals (called static windows) in order to update learning rates. Out algorithm with static windows updated the learning rates showed good performance or didn't update the learning rates even though previously updated learning rates shoved bad performance. This paper introduce a window control scheme to avoid such problems. With the window control scheme, our algorithm try to update the learning ra es only when the learning performance is continuously bad during a specified interval. If previously selected learning rates show good performance, new algorithm will not update the learning rates. This diminish the updating time of learning rates greatly. As a result, our algorithm with the window control scheme show better performance than that with static windows. In this paper, we will describe the previous and new algorithm and experimental results.

• Journal of Advanced Marine Engineering and Technology
• /
• v.22 no.5
• /
• pp.687-692
• /
• 1998

Test pattern generation is one of most difficult problems encountered in automating the design of logic circuits. The goal is to obtain the highest fault coverage with the minimum number of test patterns for a given circuit and fault set. although there have been many deterministic algorithms and heuristics the problem is still highly complex and time-consuming. Therefore new approach-es are needed to augment the existing techniques. This paper considers the problem of test pattern improvement for combinational circuits as a restricted subproblem of the test pattern generation. The problem is to maximize the fault coverage with a fixed number of test patterns for a given cir-cuit and fault set. We propose a new approach by use of a genetic algorithm. In this approach the genetic algorithm evolves test patterns to improve their fault coverage. A fault simulation is used to compute the fault coverage of the test patterns Experimental results show that the genetic algorithm based approach can achieve higher fault coverages than traditional techniques for most combinational circuits. Another advantage of the approach is that the genetic algorithm needs no detailed knowledge of faulty circuits under test.

• ETRI Journal
• /
• v.15 no.3
• /
• pp.47-59
• /
• 1994

A capacity expansion planning problem with buy-or-lease decisions is considered. Demands for capacity are deterministic and are given period-dependently at each period. Capacity additions occur by buying or leasing a capacity, and leased capacity at any period is reconverted to original source after a fixed length of periods, say, lease period. All cost functions (buying, leasing and idle costs) are assumed to be concave. And shortages of capacity and disposals are not considered. The properties of an optimal solution are characterized. This is then used in a tree search algorithm for the optimal solution and other two algorithms for a near-optimal solution are added. And these algorithms are illustrated with numerical examples.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2000.11a
• /
• pp.409-412
• /
• 2000

본 논문에서는 기존의 결정론적 볼츠만 머신의 학습알고리듬을 수정하여 양자화결합을 갖는 볼츠만 머신에도 적용할 수 있는 알고리듬을 제안하였다. 제안한 알고리듬은 2-입력 XOR문제와 3-입력 패리티문제에 적용하여 성능을 분석하였다. 그 결과 하중이 대폭적으로 양자화된 네트워크도 학습이 가능하다는 것은 은닉 뉴런수를 증가시키면 한정된 하중값의 범위로 유지할 수 있다는 것을 보여주었다. 또한 1회에 갱신하는 하중의 개수 m$_{s}$를 제어함으로써 학습계수를 제어하는 효과가 얻어지는 것을 확인하였다..

• Journal of the Korean Operations Research and Management Science Society
• /
• v.12 no.2
• /
• pp.5-20
• /
• 1987

This paper addresses sensitivity analysis for a deterministic multi-period production and inventory model. The model assumes a piecewise linear cost structure, but permits backlogging of unsatisfied demand. Our approach to sensitivity analysis here can be divided into two basic steps; (1) to find the optimal production policy through a forward dynamic programming algorithm similar to the backward version of Zangwill [1966] and (2) to apply the penalty network approach by the author [1986] in order to derive sensitivity ranges for various model parameters. Computational aspects are discussed and topics of further research are suggested.

• Proceedings of the KIEE Conference
• /
• 2007.04a
• /
• pp.3-5
• /
• 2007

This paper presents decision feedback equalizers using a recurrent neural network trained algorithm using extended Kalman filter(EKF) and sigma-point Kalman filter(SPKF). EKF is propagated, analytically through the first-order linearization of the nonlinear system. This can introduce large errors in the true posterior mean and covariance of the Gaussian random variable. The SPKF addresses this problem by using a deterministic sampling approach. The features of the proposed recurrent neural equalizer And we investigate the bit error rate(BER) between EKF and SPKF.

• 전기의세계
• /
• v.22 no.2
• /
• pp.28-32
• /
• 1973

System modeling is applied in developing a probabilistic linear estimator for the load of an electric power system for the purpose of short term load forecasting. The model assumer that the load in given by the suns of a periodic discrete time serier with a period of 24 hour and a residual term such that the output of a discrete time dynamical linear system driven by a white random process and a deterministic input. And also we have established the main forecasting algorithms, which are essemtally the Kalman filter-predictor equations.

• Proceedings of the KIEE Conference
• /
• 1988.07a
• /
• pp.902-904
• /
• 1988

For a generalized minimum variance controller algorithm the weighting polynomials are are calculated in a way to assign the closed loop poles of the system and to specify the controller gain at a frequency. As a result the oscillations in the control signal may be reduced without changing the deterministic behaviour of the system.

• Journal of Biomedical Engineering Research
• /
• v.12 no.3
• /
• pp.171-176
• /
• 1991

A solution for the course of the general deterministic epidemic model is obtained by elementary series expansion. This is valid over all times, and appears to hold accurate]y over a very wide range of population and threshould parameter values. This algorithm can be more efficient than either numerical or recursive procedures in terms of the number of operations required to evaluate a sequence of points along the course of the epidemic.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1993.06a
• /
• pp.1133-1136
• /
• 1993

Fuzzy algorithm is essentially nondeterministic, but to guarantee the stable control the fuzzy control program should be deterministic in practice. Fuzzy controllers with matrix representation is very simple in construction and very fast in computation. The value of the matrix is not adequate at the first place, but can be modified by using the neural networks. We apply the simple heuristic techniques to modify the matrix successfully.