• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.3
• /
• pp.503-511
• /
• 1995

A batch-mode algorithm is proposed to increase the speed of learning in the error backpropagation algorithm with variable learning rate and variable momentum parameters in classification problems. The objective function is normalized with respect to the number of patterns and output nodes. Also the gradient of the objective function is normalized in updating the connection weights to increase the effect of its backpropagated error. The learning rate and momentum parameters are determined from a function of the gradient norm and the number of weights. The learning rate depends on the square rott of the gradient norm while the momentum parameters depend on the gradient norm. In the two typical classification problems, simulation results demonstrate the effectiveness of the proposed algorithm.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2007.06a
• /
• pp.124-129
• /
• 2007

For seismic imaging and inversion, the inverted image depends on how we define the objective function. ${\ell}^1$-norm is more robust than ${\ell}^2$-norm. However, it is difficult to apply the Newton-type algorithm directly because the partial derivative for ${\ell^1$-norm has a singularity. In our paper, to overcome the difficulties of singularities, Huber function given by hybrid ${\ell}^1/{\ell}^2$-norm is used. We tested the robustness of our new object function with several noisy data set. Numerical results show that the new objective function is more robust to band limited spiky noise than the conventional object function.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.1
• /
• pp.68-75
• /
• 1999

Machine recognition of faces from still and video images is emerging as an active research area spanning several disciplines such as image processing, pattern recognition, computer vision and neural networks. In addition, human face identification has numerous applications such as human interface based systems and real-time video systems of surveillance and security. In this paper, we propose an algorithm that can identify a particular individual face. We consider human face identification system in color space, which hasn't often considered in conventional in conventional methods. In order to make the algorithm insensitive to luminance, we convert the conventional RGB coordinates into normalized CIE coordinates. The normalized-CIE-based facial images are KL-transformed. The transformed data are used as the input of multi-layered neural network and the network are trained using error-backpropagation methods. Finally, we verify the system performance of the proposed algorithm by experiments.

• Proceedings of the KIEE Conference
• /
• 2003.11c
• /
• pp.1010-1013
• /
• 2003

This paper describes an algorithm for determining robot location using appearance-based paradigm. This algorithm compress the image set using PCA(principal component analysis) to obtain a low-dimensional subspace, called the eigenspace, and it makes a manifold that represent a continuous-appearance function. To determine robot location, given an unknown input image, the recognition system first projects the image to eigenspace. Neural network use coefficients of the eigenspace to estimate the location of the mobile robot. The algorithm has been implemented and tested on a mobile robot system. In several trials it computes location accurately.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.1
• /
• pp.1-6
• /
• 2013

For predicting defect-prone module in software, SVM classifier showed good performance in a previous research. But there are disadvantages that SVM parameter should be chosen differently for every kernel, and algorithm should be performed iteratively for predict results of changed parameter. Therefore, we find these parameters using Genetic Algorithm and compare with result of classification by Backpropagation Algorithm. As a result, the performance of GA-SVM model is better.

• The Journal of Korea Robotics Society
• /
• v.6 no.4
• /
• pp.301-307
• /
• 2011

Based on the stability criteria of ZMP (Zero Moment Point), this paper proposes an adjusting algorithm that modifies walking trajectory of a bipedal robot for stable walking by analyzing ZMP trajectory of it. In order to maintain walking balance of the bipedal robot, ZMP should be located within a supporting polygon that is determined by the foot supporting area with stability margin. Initially tilting imposed to the trajectory of the upper body is proposed to transfer ZMP of the given walking trajectory into the stable region for the minimum stability. A neural network method is also proposed for the stable walking trajectory of the biped robot. It uses backpropagation learning with angles and angular velocities of all joints with tilting to get the improved walking trajectory. By applying the optimized walking trajectory that is obtained with the neural network model, the ZMP trajectory of the bipedal robot is certainly located within a stable area of the supporting polygon. Experimental results show that the optimally learned trajectory with neural network gives more stability even though the tilting of the pelvic joint has a great role for walking stability.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.04b
• /
• pp.616-618
• /
• 2002

MLP(multilayer perceptron)는 다른 패턴인식 방법에 비해 몇 가지 이점이 있어 다양한 문제영역에서 사용되고 있다 그러나 MLP의 학습에 일반적으로 사용되는 EBP(error backpropagation) 알고리즘은 학습시간이 비교적 오래 걸린다는 단점이 있으며, 이는 실시간 처리를 요구하는 문제나 대규모 데이터 및 MLP 구조로 인해 학습시간이 상당히 긴 문제에서 제약으로 작용한다. 패턴인식에 사용되는 학습데이터는 풍부한 중복특성을 내포하고 있으므로 패턴마다 MLP의 내부변수를 갱신하는 은라인 계열의 학습방식이 속도의 향상에 상당한 효과가 있다. 일반적인 온라인 EBP 알고리즘에서는 내부 가중치 갱신시 고정된 학습률을 적용한다. 고정 학습률을 적절히 선택함으로써 패턴인식 응용에서 상당한 속도개선을 얻을 수 있지만, 학습률을 고정함으로써 온라인 방식에서 패턴별 갱신의 특성을 완전히 활용하지 못하는 비효율성이 발생한다. 또한, 학습도중 패턴군이 학습된 패턴과 그렇지 못한 패턴으로 나뉘고 이 가운데 학습된 패턴은 학습을 위한 계산에 포함될 필요가 없음에도 불구하고, 기존의 온라인 EBP에서는 에폭에 할당된 모든 패턴을 일률적으로 계산에 포함시킨다. 이 문제에 대해 본 논문에서는 학습이 진행됨에 따라 패턴마다 적절한 학습률을 적용하고 필요한 패턴만을 학습에 반영하는 패턴별 가변학습률 및 학습생략(COIL) 방댑을 제안한다. 제안한 COIL의 성능을 입증하기 위해 화자증명과 음성인식을 실험하고 그 결과를 제시한다.

• Structural Engineering and Mechanics
• /
• v.21 no.2
• /
• pp.205-220
• /
• 2005

This paper presents the possibilities of adapting artificial neural networks (ANNs) to predict the dimensionless parameters related to the maximum contact pressures of an elasticity problem. The plane symmetric double receding contact problem for a rigid stamp and two elastic strips having different elastic constants and heights is considered. The external load is applied to the upper elastic strip by means of a rigid stamp and the lower elastic strip is bonded to a rigid support. The problem is solved under the assumptions that the contact between two elastic strips also between the rigid stamp and the upper elastic strip are frictionless, the effect of gravity force is neglected and only compressive normal tractions can be transmitted through the interfaces. A three layered ANN with backpropagation (BP) algorithm is utilized for prediction of the dimensionless parameters related to the maximum contact pressures. Training and testing patterns are formed by using the theory of elasticity with integral transformation technique. ANN predictions and theoretical solutions are compared and seen that ANN predictions are quite close to the theoretical solutions. It is demonstrated that ANNs is a suitable numerical tool and if properly used, can reduce time consumed.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.50 no.4
• /
• pp.201-207
• /
• 2001

This paper describes an automated detection algorithm of ST-T episodes using global curvature which can connect the isoelectric level in ECG and can eliminate not only the slope of ST segment, but also difference of the baseline and global curve. This above method of baseline correction is very faster than the classical baseline correction methods. The optimal values of parameters for baseline correction were found as the value having the highest detection rate of ST episode. The features as input of backpropagation Neural Network were extracted from the whole ST segment. The European ST-T database was used as training and test data. Finally, ST elevation, ST depression and normal ST were classified. The average ST episode sensitivity and predictivity were 85.42%, 80.29%, respectively. This result shows the high speed and reliability in ST episode detection. In conclusion, the proposed method showed the possibility in various applications for the Holter system.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.10 no.6
• /
• pp.1032-1037
• /
• 2006

The channel is one of the precious and limited resources in wireless networks. There are many researches on the channel management. Recently, the optimization problem of guard channels has been an important issue. In this paper, we propose an intelligent channel management technique based on the neural networks. An SRN channel allocation model is developed to generate the learning data for the neural networks and the performance analysis of system. In the proposed technique, the neural network is trained to generate optimal guard channel number g, using backpropagation supervised learning algorithm. The optimal g is computed using the neural network and compared to the g computed by the SRM model. The numerical results show that the difference between the value of 8 by backpropagation and that value by SRM model is ignorable.