• Proceedings of the Korean Society for Rock Mechanics Conference
• /
• 1996.03a
• /
• pp.35-42
• /
• 1996

지하 구조물의 안정성 확보와 경제적인 시공을 위하여 상세하고 합리적인 암반분류가 필요하다. 설계 초기에는 제한적인 정보와 암반의 불확실성에 따라 암반분류의 신뢰도가 떨어진다. 이러한 불확실한 지질 정보를 근사하게 추론할 수 있는 방법으로서 인공지능(Artificial intelligence) 특히 인공신경망 (Artificial neural network)이 있다. (중략)

• Journal of the Korean Geotechnical Society
• /
• v.20 no.8
• /
• pp.77-87
• /
• 2004

In this paper a back-propagation neural network model is developed to estimate the preconsolidation pressure of Korean soft soils based on 176 oedometer tests and 63 piezocone test results, which were compiled from 11 sites - western and southern parts of Korea. Only 147 data were used for the training of the neural network and 29 data, which were not used during the training phase, were used for the verification of trained network. Empirical and theoretical models were compared with the developed neural network model. A simple 4-4-9-1 multi-layered neural network has been developed. The cone tip resistance $q_T$ penetration pore pressure $u_2$, total overburden pressure $\sigma_{vo}$ and effective overburden pressure $\sigma'_{vo}$ were selected as input variables. The developed neural network model was validated by comparing the prediction results of the proposed neural network model for the new data which were not used for the training of the model with the measured preconsolidation pressures. It can also predict more precise and reliable preconsolidation pressures than the analytical and empirical model. Furthermore, it can be carefully concluded that neural network model can be used as a generalized model for prediction of preconsolidation pressure throughout Korea since developed model shows good performance for the new data which were not used in both training and testing data.

• Journal of the Korean Geotechnical Society
• /
• v.27 no.1
• /
• pp.17-24
• /
• 2011

The characteristics of flow and strength of CLSM depend on the combination ratio including the fly ash, pond ash, cement, water quantity and etc. However, it is very difficult to draw the mechanism about the flow, strength and the mixing ratio of each components. Therefore, the method of calculation drawing the flow about the component ratio of CLSM and compression strength value is needed for the valid practical use of CLSM. To verify the efficiency of artificial neural network, new data which were not used for establishing the model were predicted and compared with the results of laboratory tests. In this research, it was used to evaluate the learning efficiency of the artificial neural network model and the prediction ability by changing the node number of hidden layer, learning rate, momentum, target system error and hidden layer. By using the results, the optimized artificial neural network model which is suitable for a flow and compressive strength estimate of CLSM was determined.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.23 no.1
• /
• pp.109-117
• /
• 2011

It was calculated using empirical formulas for the weight of Tetrapod, which was a representative armor unit in the rubble mound breakwater in Korea. As the formulas were evaluated from a curve-fitting with the result of hydraulic test, the uncertainty of experimental error was included. Therefore, the neural network and M5' model tree were used to minimize the uncertainty and predicted the stability number of armor block. The index of agreement between the predicted and measured stability number was calculated to assess the degree of uncertainty for each model. While the neural network with the highest index of agreement have an excellent prediction capability, a significant disadvantage exists that general designers can not easily handle the method. However, although M5' model tree has a lower prediction capability than the neural network, the model tree is easily used by the designers because it has a good prediction capability compared with the existing empirical formula and can be used to propose the formulas like an empirical formula.

• Proceedings of the Optical Society of Korea Conference
• /
• 1991.07a
• /
• pp.55-59
• /
• 1991

신경회로망은 뒤뇌의 신경조직이 갖는 병렬적이며 분산적인 정보처리 능력을 흉내낸 인공적인 회로망이다. 이러한 신경회로망을 영상인식, 음성인식, 적응제어 및 최적화등에 응용할 경우 지금까지 얻지 못하였던 우수한 여러 가지 특성을 얻을수 있음을 알려짐에 따라 신경회로망을 구체적으로 구현하고자 하는 연구가 활발히 이루어지고 있다. 본 고에서는 신경소자간의 연결세기의 변조에 의한 학습 원리를 설명하고 광전기적인 그현방법에 대해서 몇 개의 예를 들어 설명하고 그 발전 가능성에 대하여 기술하였다.

• Journal of the Korean Chemical Society
• /
• v.44 no.2
• /
• pp.102-108
• /
• 2000

Relative retention times (RRTs) of RAH molecules in HPLC are trained and predicted intesting sets using a multiple linear regression (NLR) and an artificial neural network (ANN). The maindescriptors in QSRR are molecular connectivity ($^1X_v,\;^2X_v$), the length-to-breadth ratios (L/B), and molecular dipole moment(D). L/B which is related with slot model is a good descripter in ANN, but isn't in MLR. Varainces which show the accuracy of prediction times in testing sets are 0.0099, 0.0114 for ANN and MLR, respectively. It was shown that ANN can exceed the MLR in prediction accuracy.

• Journal of Broadcast Engineering
• /
• v.24 no.1
• /
• pp.87-96
• /
• 2019

Improvement of low-light image mainly focuses on the contrast enhancement. Many researches have been carried out for brightness enhancement, contrast improvement and illumination reduction. Recently, the aforementioned approaches have been replaced by artificial neural networks. This paper proposes a methodology that can replace the Retinex-based reflectance image acquisition by deep neural network. Experiments carried out on 102 low-light images validated the feasibility of the replacement by producing PSNR=30.8682(db) and SSIM=0.4345.

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.441-447
• /
• 1999

본 논문은 복잡 적응 시스템의 분석 및 모델링을 위해, 인공생명의 기본 패러다임인 셀룰라 오토마타를 선택하여, 무정형의 구조를 가지며 투명한 자료 전파 특성을 갖는 셀룰라 신경 회로망의 설계하고 개발하는데 중점을 두었다. 우선, 신경 회로망의 불규칙한 구조를 발생학적으로 다루어 무정형의 은닉층을 생성하고, 다윈의 진화론을 적용하여 구조적 진화 및 선택을 통해 최적화된 신경 회로망을 설계하였다. 주변 셀의 상태를 감지하여 자신의 상태를 수정해나가는 방식의 셀룰라 오토마타의 투명한 신호 전파 모델로 자료 및 오차의 역전파에 적용하도록 고안하였고, 라마르크의 용불용설을 활용한 오차의역전파 학습 알고리즘을 유도하였다. 이러한 복잡 적응계의 학습 과정을 유도하여 시뮬레이션에서 그 타당성을 입증하였다. 시뮬레이션에서는 신경 회로망의 XOR 문제와 다중 입력 다중 출력 함수에 대한 근사화 문제를 풀었다.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2022.11a
• /
• pp.388-389
• /
• 2022

Recently, AI (Artificial Intelligence) has been applied to various technologies such as automatic driving, robot and smart communication. Currently, AI system is developed by software-based method using tensor flow, and GPU (Graphic Processing Unit) is employed for processing unit. However, if software-based method employing GPU is used for AI applications, there is a problem that we can not change the internal circuit of processing unit. In this method, if high-level jobs are required for AI system, we need high-performance GPU, therefore, we have to change GPU or graphic card to perform the jobs. In this work, we developed a CNN-based FPGA (Field Programmable Gate Array) chip to solve this problem.

• KIISE Transactions on Computing Practices
• /
• v.23 no.5
• /
• pp.322-327
• /
• 2017

The classification problem in the field of Natural Language Processing has been studied for a long time. Continuing forward with our previous research, which classifies large-scale text using Convolutional Neural Networks (CNN), we implemented Recurrent Neural Networks (RNN), Long-Short Term Memory (LSTM) and Gated Recurrent Units (GRU). The experiment's result revealed that the performance of classification algorithms was Multinomial Naïve Bayesian Classifier < Support Vector Machine (SVM) < LSTM < CNN < GRU, in order. The result can be interpreted as follows: First, the result of CNN was better than LSTM. Therefore, the text classification problem might be related more to feature extraction problem than to natural language understanding problems. Second, judging from the results the GRU showed better performance in feature extraction than LSTM. Finally, the result that the GRU was better than CNN implies that text classification algorithms should consider feature extraction and sequential information. We presented the results of fine-tuning in deep neural networks to provide some intuition regard natural language processing to future researchers.