• Journal of the Korea Society for Simulation
• /
• v.18 no.2
• /
• pp.49-55
• /
• 2009

In this paper, we investigate an optimal allocation of constant service resources in stochastic system to optimize the expected performance of interest. For this purpose, we use the control variates to estimate the gradients of expected performance with respect to given resource parameters, and apply these estimated gradients in stochastic optimization algorithm to find the optimal allocation of resources. The proposed gradient estimation method is advantageous in that it uses simulation results of a single design point without increasing the number of design points in simulation experiments and does not need to describe the logical relationship among realized performance of interest and perturbations in input parameters. We consider the applications of this research to various models and extension of input parameter space as the future research.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.30 no.3C
• /
• pp.141-146
• /
• 2005

The authors investigate the convergence speed problem of nonlinear adaptive equalization. Convergence constraints and time constant of radial basis function network using stochastic gradient (RBF-SG) algorithm is analyzed and a method of making time constant independent of hidden-node output power by using sample-by-sample node output power estimation is derived. The method for estimating the node power is to use a single-pole low-pass filter. It is shown by simulation that the proposed algorithm gives faster convergence and lower minimum MSE than the RBF-SG algorithm.

• Journal of Korean Society of Transportation
• /
• v.24 no.4 s.90
• /
• pp.149-159
• /
• 2006

Traditional trip tables are estimated through large-scale surveys such as household survey, roadside interviews, and license Plate matching. These methods are, however, expensive and time consuming. This paper presents two origin-destination (OD) trip matrix estimation methods from link traffic counts in stochastic assignment, which contains perceived errors of drivers for alternatives. The methods are formulated based on the relation between link flows and OD demands in logit formula. The first method can be expressed to minimize the difference between observed link flows and estimated flows, derived from traffic assignment and be solved by gradient method. The second method can be formulated based on dynamic process, which nay describe the daily movement patterns of drivers and be solved by a recursive equation. A numerical example is used for assessing the methods, and shows the performances and properties of the models.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.8
• /
• pp.777-782
• /
• 2011

Neural network technique is widely employed in the fields of signal processing, control systems, pattern recognition, etc. Learning of neural networks is an important procedure to accomplish dynamic system modeling. This paper presents a novel learning approach for differential neural network models based on the Kalman-Bucy filter theory. We construct an augmented state vector including original neural state and parameter vectors and derive a state estimation rule avoiding gradient function terms which involve to the conventional neural learning methods such as a back-propagation approach. We carry out numerical simulation to evaluate the proposed learning approach in nonlinear system modeling. By comparing to the well-known back-propagation approach and Kalman-Bucy filtering, its superiority is additionally proved under stochastic system environments.

• Journal of the Korean Data and Information Science Society
• /
• v.7 no.1
• /
• pp.113-118
• /
• 1996

Principal component analysis(PCA) is an essential technique for data compression and feature extraction, and has been widely used in statistical data analysis, communication theory, pattern recognition, and image processing. Oja(1992) found that a linear neuron with constrained Hebbian learning rule can extract the principal component by using stochastic gradient ascent method. In practice real data often contain some outliers. These outliers will significantly deteriorate the performances of the PCA algorithms. In order to make PCA robust, Xu & Yuille(1995) applied statistical physics to the problem of robust principal component analysis(RPCA). Devlin et.al(1981) obtained principal components by using techniques such as M-estimation. The propose of this paper is to investigate from the statistical point of view how Xu & Yuille's(1995) RPCA works under the same simulation condition as in Devlin et.al(1981).

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.58-58
• /
• 2019

최근 기계학습기법에 대한 활발한 연구로 인하여 많은 기계학습기법들이 개발되었다. 이러한 최신기계학습기법은 기존에 사용되어온 기계학습기법과 경험식들보다 자연현상을 예측하고 재현하는데 높은 성능을 보이는 것으로 알려져 있다. 레이더 자료를 이용한 강우추정 기법으로는 ZR관계식이 널리 사용되고 있다. 이상적인 조건에서는 ZR 관계식을 이용한 레이더 강우추정이 양호한 성능을 보이나, 실제 레이더 자료를 이용한 강우추정은 이상적인 환경이 아닌 경우가 매우 많다. 이런 ZR관계식의 한계점을 보완하기 위한 방법으로 기계학습기법을 이용한 레이더 강우추정 기법들이 개발되었으나, 현재 한국의 레이더 자료를 대상으로 해서는 많은 연구가 진행되어 오지 않고 있다. 레이더 자료를 이용한 강우추정의 정확도 향상을 위해서는 최신 기계학습기법들의 레이더 강우추정 기법에 대한 적용가능성을 평가해 볼 필요성이 있다. 본 연구에서는 random forest, stochastic gradient boosted model, extreme learning machine의 강우 레이더 강우추정 기법으로의 적용성을 평가하였다. 강우추정 기법 개발 및 성능 비교를 위해서 2018년 광덕산 이중편파 레이더 자료를 이용하였다. 다양한 이중편파 매개변수 조합을 레이더 강우추정 기법의 입력변수로 적용하였다. 기존 연구의 사용되어 온 ZR관계식의 매개변수를 또한 강우사상과 이중편파 매개변수 조합을 이용하여 추정하였다. 기계학습을 적용한 레이더 강우추정 기법이 ZR관계식보다 상관계수와 제곱근오차를 기준으로 높은 강우추정 정확도를 보였다. 특히 개발된 강우추정 기법은 호우사상에서 높은 정확도를 보이는 것을 확인 할 수 있었다. 적용된 기계학습 기법 중에서는extreme learning machine이 레이더 강우추정기법 개발에 가장 적합한 것으로 나타났다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.10
• /
• pp.2310-2316
• /
• 1997

This paper presents a comparative performance analysis of the stochastic gradient adaptive algorithm based on the least mean absolute third (LMAT) error criterion with other widely-used competing adaptive algorithms. Under the assumption that the signals involved are zero-mean, wide-sense stationary and Gaussian, approximate expressions that characterize the steady-state mean-squared estimation error of the algorithm is dervied. The validity of our derivation is then confirement by computer simulations. The convergence speed is compared under the condition that the LMAT and other competing algorithms converge to the same value for the mean-squared estimation error in the stead-state, and superior convergence property of the LMAT algorithm is observed. In particular, it is shown that the LMAT algorithm converges faster than other algorithms even through the eignevalue spread ratio of the input signal and measurement noise power change.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1329-1341
• /
• 2022

Destroying the marine environment and marine ecosystem and causing marine accidents, marine debris is generated every year, and among them, submerged marine debris is difficult to identify and collect because it is on the seabed. Therefore, deep-learning-based semantic segmentation was experimented on waste fish nets and waste ropes using underwater images to identify efficient collection and distribution. For segmentation, a high-resolution network (HRNet), a state-of-the-art deep learning technique, was used, and the performance of each optimizer was compared. In the segmentation result fish net, F1 score=(86.46%, 86.20%, 85.29%), IoU=(76.15%, 75.74%, 74.36%), For the rope F1 score=(80.49%, 80.48%, 77.86%), IoU=(67.35%, 67.33%, 63.75%) in the order of adaptive moment estimation (Adam), Momentum, and stochastic gradient descent (SGD). Adam's results were the highest in both fish net and rope. Through the research results, the evaluation of segmentation performance for each optimizer and the possibility of segmentation of marine debris in the latest deep learning technique were confirmed. Accordingly, it is judged that by applying the latest deep learning technique to the identification of submerged marine debris through underwater images, it will be helpful in estimating the distribution of marine sedimentation debris through more accurate and efficient identification than identification through the naked eye.

• Structural Engineering and Mechanics
• /
• v.32 no.1
• /
• pp.95-126
• /
• 2009

The present contribution addresses the parallelization of advanced simulation methods for structural reliability analysis, which have recently been developed for large-scale structures with a high number of uncertain parameters. In particular, the Line Sampling method and the Subset Simulation method are considered. The proposed parallel algorithms exploit the parallelism associated with the possibility to simultaneously perform independent FE analyses. For the Line Sampling method a parallelization scheme is proposed both for the actual sampling process, and for the statistical gradient estimation method used to identify the so-called important direction of the Line Sampling scheme. Two parallelization strategies are investigated for the Subset Simulation method: the first one consists in the embarrassingly parallel advancement of distinct Markov chains; in this case the speedup is bounded by the number of chains advanced simultaneously. The second parallel Subset Simulation algorithm utilizes the concept of speculative computing. Speedup measurements in context with the FE model of a multistory building (24,000 DOFs) show the reduction of the wall-clock time to a very viable amount (<10 minutes for Line Sampling and ${\approx}$ 1 hour for Subset Simulation). The measurements, conducted on clusters of multi-core nodes, also indicate a strong sensitivity of the parallel performance to the load level of the nodes, in terms of the number of simultaneously used cores. This performance degradation is related to memory bottlenecks during the modal analysis required during each FE analysis.

• Structural Monitoring and Maintenance
• /
• v.5 no.1
• /
• pp.51-65
• /
• 2018

A number of sensing techniques have been implemented for detecting defects in civil infrastructures instead of onsite human inspections in structural health monitoring. However, the issue of faults in sensors has not received much attention. This issue may lead to incorrect interpretation of data and false alarms. To overcome these challenges, this article presents a deep learning-based method with a new architecture of Stateful Long Short Term Memory Neural Networks (S-LSTM NN) for detecting sensor fault without going into details of the fault features. As LSTMs are capable of learning data features automatically, and the proposed method works without an accurate mathematical model. The detection of four types of sensor faults are studied in this paper. Non-stationary acceleration responses of a three-span continuous bridge when under operational conditions are studied. A deep network model is applied to the measured bridge data with estimation to detect the sensor fault. Another set of sensor output data is used to supervise the network parameters and backpropagation algorithm to fine tune the parameters to establish a deep self-coding network model. The response residuals between the true value and the predicted value of the deep S-LSTM network was statistically analyzed to determine the fault threshold of sensor. Experimental study with a cable-stayed bridge further indicated that the proposed method is robust in the detection of the sensor fault.