• Journal of the Korea Society of Computer and Information
• /
• v.4 no.2
• /
• pp.27-38
• /
• 1999

This paper proposes a high speed digital neuron processor which uses the residue number system, making the high speed operation possible without carry propagation,. Consisting of the MAC(Multiplier and with Accumulator) operation unit, quotient operation unit and sigmoid function operation unit, the neuron processor is designed through 0.8$\mu$m CMOS fabrication. The result shows that the new implemented neuron processor can run at the speed of 19.2 nSec and the size can be reduced to 1/2 compared to the neuron processor implemented by the real number operation unit.

• Proceedings of the IEEK Conference
• /
• 2009.05a
• /
• pp.405-407
• /
• 2009

SVM(Support Vector Machine) is a classification method which is recently watched in mechanical learning system. Vapnik, Osuna, Platt etc. had suggested methodology in order to solve needed QP(Quadratic Programming) to realize SVM so that have extended application field. SVM find hyperplane which classify into 2 class by converting from input space converter vector to characteristic space vector using Kernel Function. This is very systematic and theoretical more than neural network which is experiential study method. Although SVM has superior generalization characteristic, it depends on Kernel Function. There are three category in the Kernel Function as Polynomial Kernel, RBF(Radial Basis Function) Kernel, Sigmoid Kernel. This paper has analyzed performance of SVM against kernel using virtual data.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.10
• /
• pp.75-80
• /
• 2017

In this paper, we describe the effect of the regularization method and the network with identity mapping on the performance of the activation functions in deep convolutional neural networks. The activation functions act as nonlinear transformation. In early convolutional neural networks, a sigmoid function was used. To overcome the problem of the existing activation functions such as gradient vanishing, various activation functions were developed such as ReLU, Leaky ReLU, parametric ReLU, and ELU. To solve the overfitting problem, regularization methods such as dropout and batch normalization were developed on the sidelines of the activation functions. Additionally, data augmentation is usually applied to deep learning to avoid overfitting. The activation functions mentioned above have different characteristics, but the new regularization method and the network with identity mapping were validated only using ReLU. Therefore, we have experimentally shown the effect of the regularization method and the network with identity mapping on the performance of the activation functions. Through this analysis, we have presented the tendency of the performance of activation functions according to regularization and identity mapping. These results will reduce the number of training trials to find the best activation function.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.2
• /
• pp.381-389
• /
• 1996

A simple modification of the standard back-propagation algorithm to eliminate redundant connections(weights and biases) is described. It was motivated by speculations from the distribution of the magnitudes of the weights and the biases, analysis of the classification boundary, and the nonlinearity of the sigmoid function. After initial training, this algorithm eliminates all connections of which magnitude is below a threshold by setting them to zero. The algorithm then conducts retraining in which all weights and biases are adjusted to allow important ones to recover. In studies with Boolean functions, the algorithm reconstructed the theoretical minimum architecture and eliminated the connections which are not necessary to solve the functions. For simulated random signal classification problems, the algorithm produced the result which is consistent with the idea that easier problems require simpler networks and yield lower misclassification rates. Furthermore, in comparison, our algorithm produced better generalization than the standard algorithm by reducing over fitting and pattern memorization problems.

• The Journal of Society for e-Business Studies
• /
• v.15 no.1
• /
• pp.157-171
• /
• 2010

To effectively maintain manufacturing collaboration, the development of effective performance measurements for the collaboration is required. Most existing key performance indicators however were developed to measure the performances of internal activities or outsourcing of a company. For that reason, it is necessary to devise new key performance indicators that the partners participating in the collaboration can arrange and compromise with each other to reflect their common goals. In this paper, we propose collaborative Key Performance Indicators(cKPIs), which is used to measure the collaboration work of multiple manufacturing partners on the basis of the Supply Chain Operations Reference(SCOR) model. Also, a modified Sigmoid function is devised as a desirability function to reflect the characteristics of Service Level Agreement(SLA) between two partners. The proposed indicators and the desirability functions can be utilized to perform and maintain the successful collaboration by providing a way to the quantitative measurement.

• Journal of Broadcast Engineering
• /
• v.24 no.2
• /
• pp.234-242
• /
• 2019

In this paper, we propose a deep neural network which extracts SIFT feature points by determining whether the center pixel of a cropped image is a SIFT feature point. The data set of this network consists of a DIV2K dataset cut into $33{\times}33$ size and uses RGB image unlike SIFT which uses black and white image. The ground truth consists of the RobHess SIFT features extracted by setting the octave (scale) to 0, the sigma to 1.6, and the intervals to 3. Based on the VGG-16, we construct an increasingly deep network of 13 to 23 and 33 convolution layers, and experiment with changing the method of increasing the image scale. The result of using the sigmoid function as the activation function of the output layer is compared with the result using the softmax function. Experimental results show that the proposed network not only has more than 99% extraction accuracy but also has high extraction repeatability for distorted images.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.15 no.2
• /
• pp.1109-1117
• /
• 2014

We study free vibration analysis of sigmoid functionally graded materials(S-FGM) nano-scale plates, using a nonlocal elasticity theory of Eringen in this paper. This theory has ability to capture the both small scale effects and sigmoid function in terms of the volume fraction of the constituents for material properties through the plate thickness. Numerical solutions of S-FGM nano-scale plate are presented using this theory to illustrate the effect of nonlocal theory on natural frequency of the S-FGM nano-scale plates. The relations between nonlocal and local theories are discussed by numerical results. Further, effects of (i) power law index (ii) nonlocal parameters, (iii) elastic modulus ratio and (iv) thickness and aspect ratios on nondimensional frequencies are investigated. In order to validate the present solutions, the reference solutions are compared and discussed. The results of S-FGM nano-scale plates using the nonlocal theory may be the benchmark test for the free vibration analysis.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.8
• /
• pp.17-22
• /
• 2016

For the self-adaptive equalization of higher-order QAM systems, this paper proposes a new soft decision-directed (SDD) algorithm that opens the eye patterns quickly as well as significantly reducing the error level in the steady-state when it is applied to the initial equalization stage with completely closed eye patterns. The proposed method for M-QAM application minimized the computational complexity of the existing SDD by the symbol estimated based on the two symbols closest to the observation, and greatly simplified the soft decision independently of the QAM order. Furthermore, in the symbol estimating it increased the reliability of the estimates by applying the superior properties of the sigmoid function and avoiding the erroneous estimation of the threshold function. In addition, the initialization performance was improved when an error is generated to update the equalizer, weighting the symbol decision by the threshold function to the error, resulting in an extension of the range of error fluctuations. As a result, the proposed method improves remarkably the computational complexity and the properties of initialization and convergence of the traditional SDD. Through simulations for 64-QAM and 256-QAM under multipath channel conditions with additive noise, the usefulness of the proposed methods was confirmed by comparing the performance of the proposed 2-SDD and two forms of weighted 2-SDD with CMA.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.363-363
• /
• 2019

활성화 함수(activation function)는 기계학습(machine learning)의 학습과정에 비선형성을 도입하여 심층적인 학습을 용이하게 하고 예측의 정확도를 높이는 중요한 요소 중 하나이다(Roy et al., 2019). 일반적으로 기계학습에서 사용되고 있는 활성화 함수의 종류에는 계단 함수(step function), 시그모이드 함수(sigmoid 함수), 쌍곡 탄젠트 함수(hyperbolic tangent function), ReLU 함수(Rectified Linear Unit function) 등이 있으며, 예측의 정확도 향상을 위하여 다양한 형태의 활성화 함수가 제시되고 있다. 본 연구에서는 기계학습을 통하여 수위예측 시 정확도 향상을 위하여 Hybrid 활성화 함수를 제안하였다. 연구대상지는 조수간만의 영향을 받는 한강을 대상으로 선정하였으며, 2009년 ~ 2018년까지 10년간의 수문자료를 활용하였다. 수위예측 알고리즘은 Python 내 Tensorflow의 RNN (Recurrent Neural Networks) 모델을 이용하였으며, 강수량, 수위, 조위, 댐 방류량, 하천 유량의 수문자료를 학습시켜 3시간 및 6시간 후의 수위를 예측하였다. 예측정확도 향상을 위하여 입력 데이터는 정규화(Normalization)를 시켰으며, 민감도 분석을 통하여 신경망모델의 은닉층 개수, 학습률의 최적 값을 도출하였다. Hybrid 활성화 함수는 쌍곡 탄젠트 함수와 ReLU 함수를 혼합한 형태로 각각의 가중치($w_1,w_2,w_1+w_2=1$)를 변경하여 정확도를 평가하였다. 그 결과 가중치의 비($w_1/w_2$)에 따라서 예측 결과의 RMSE(Roote Mean Square Error)가 최소가 되고 NSE (Nash-Sutcliffe model Efficiency coefficient)가 최대가 되는 지점과 Peak 수위의 예측정확도가 최대가 되는 지점을 확인할 수 있었다. 본 연구는 현재 Data modeling을 통한 수위예측의 정확도 향상을 위해 기초가 되는 연구이나, 향후 다양한 형태의 활성화 함수를 제안하여 정확도를 향상시킨다면 예측 결과를 통하여 침수예보에 대한 의사결정이 가능할 것으로 기대된다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.04b
• /
• pp.586-588
• /
• 2001

최근에는 순환 신경망의 생성모델이 비교사 학습에 관련하여 활발히 연구되고 있다. 이러한 형태의 신경망은 형태 추출이나 인식에 효과적으로 사용될 수 있는 반면 반복 loop를 사용하므로 대단히 많은 계산이 필요하다. 본 논문에서는 Oh와 Seung에 의해 제안된 상향전파(Up-propagation) network이라는 순환 신경망을 FPGA를 이용해서 구현하였다. 단층 신경망은 9개의 상층 neuron과 256개의 하층 neuron으로 구성되 있으며 4만 게이트의 FPGA 하나로 효과적으로 구현할 수 있다. pipeline된 곱셈기로 게산 속도를 향상시켰고 sigmoid 전달 함수는 유한 정밀도의 2차 다항식으로 근사될 수 있다. 구현된 하드웨어는 hand-written 숫자 영상인 USPS data를 재생하는데 사용되었으며 좋은 결과를 얻었다.