• Proceedings of the KIEE Conference
• /
• 1993.07a
• /
• pp.244-247
• /
• 1993

In this paper, Fuzzy-Neural Network Control system that has the characteristic of fuzzy control to be controlled easily end the good characteristic of a artificial neural network to control the plant due to its learning is presented. A fuzzy rule to be applied is selected automatically by the allocated neurons. The neurons correspond to Fuzzy rules which ere created by a expert. To adaptivity, the more precise modeling is implemented by error beck-propagation learning of adjusting the link-weight of fuzzy membership function in Fuzzy-Neural Network. The more classified fuzzy rule is used to include the property of Dual Mode Method. To test the effectiveness of the algorithm presented above, the simulation for position end velocity of DC servo motor is implemented.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2000.07a
• /
• pp.59-62
• /
• 2000

A VLSI implementation of a self-learning neural network integrated circuits using a linearity synapse transistor is investigated. The thickness dependence of oxide current density, stress current, transient current and channel current has been measured in oxides with thicknesses between 41 and 112 $\AA$, which have the channel width$\times$length 10$\times$1${\mu}{\textrm}{m}$ respectively. The transient current will affect data retention in synapse transistors and the stress current is used to estimate to fundamental limitations on oxide thicknesses. The synapse transistor has represented the neural states and the manipulation which gave unipolar weights. The weight value of synapse transistor was caused by the bias conditions. Excitatory state and inhitory state according to weighted values affected the drain source current.

• Journal of Chest Surgery
• /
• v.18 no.4
• /
• pp.598-604
• /
• 1985

One hundred consecutive patients with the Tetralogy of Fallot underwent total correction at National Medical Center during the period 1977 to 1984, Oct. During this study period, we adopted more active policy towards reconstruction of right ventricular outflow tract across pulmonary valve. The mortality was 48% for patients less than 15 kg and 19% in patients above 15kg. Initially Bretschneider`s solution was used as cardioplegia, which was replaced by St. Thomas` solution since 1983, Jan. After then overall mortality dropped to 9% compared to 45% of initial learning period. Heart block occurred In 11 patients, 10 of whom died of combined low cardiac output syndrome. Pure low cardiac output syndrome was noted in 18 patients, most of whom responded to medical measures well except 4 patients. Recently sepsis of Serratia marcescence, which occurred explosively during several months to open heart surgery patients, attacked 3 tetralogy patients resulting in 2 hospital deaths. Our experience has shown that body weight, choice of cardioplegia and accumulation of experience as well as advance of operative and postoperative techniques are still important factors affecting survival rate at initial learning period.

• Steel and Composite Structures
• /
• v.46 no.1
• /
• pp.107-114
• /
• 2023

A new intelligent adaptive control scheme was proposed that combines observer disturbance-based adaptive control and fuzzy adaptive control for a composite structure with a mass-adjustable damper. The most important advantage is that the control structures do not need to know the uncertainty limits and the interference effect is eliminated. Three adjustable parameters in LMI are used to control the gain of the 2D fuzzy control. Binary performance indices with weighted matrices are constructed to separately evaluate validation and training performance using the revalidation learning function. Determining the appropriate weight matrix balances control and learning efficiency and prevents large gains in control. It is proved that the stability of the control system can be ensured by a linear matrix theory of equality based on Lyapunov's theory. Simulation results show that the multilevel simulation approach combines accuracy with high computational efficiency. The M-TMD system, by slightly reducing critical joint load amplitudes, can significantly improve the overall response of an uncontrolled structure.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.557-560
• /
• 2023

In this paper, we introduce a pre-training method leveraging the capabilities of the Vision Transformer (ViT) for disease diagnosis in conventional Fundus images. Recognizing the need for effective representation learning in medical images, our method combines the Vision Transformer with a Masked Autoencoder to generate meaningful and pertinent image augmentations. During pre-training, the Masked Autoencoder produces an altered version of the original image, which serves as a positive pair. The Vision Transformer then employs contrastive learning techniques with this image pair to refine its weight parameters. Our experiments demonstrate that this dual-model approach harnesses the strengths of both the ViT and the Masked Autoencoder, resulting in robust and clinically relevant feature embeddings. Preliminary results suggest significant improvements in diagnostic accuracy, underscoring the potential of our methodology in enhancing automated disease diagnosis in fundus imaging.

• Journal of the Korea Convergence Society
• /
• v.9 no.10
• /
• pp.21-28
• /
• 2018

To help users who are experiencing difficulties finding the right learning course corresponding to their level of proficiency, we developed a recommendation model for personalized learning course for Intelligence Tutoring System(ITS). The Personalized Learning Course Recommendation model for ITS analyzes the learner profile and extracts the keyword by calculating the weight of each word. The similarity of vector between extracted words is measured through the cosine similarity method. Finally, the three courses of top similarity are recommended for learners. To analyze the effects of the recommendation model, we applied the recommendation model to the Women's ability development center. And mean, standard deviation, skewness, and kurtosis values of question items were calculated through the satisfaction survey. The results of the experiment showed high satisfaction levels in accuracy, novelty, self-reference and usefulness, which proved the effectiveness of the recommendation model. This study is meaningful in the sense that it suggested a learner-centered recommendation system based on machine learning, which has not been researched enough both in domestic, foreign domains.

• The Journal of the Korea Contents Association
• /
• v.18 no.11
• /
• pp.447-454
• /
• 2018

This paper proposes a depth image generation algorithm of stereo images using a deep learning model composed of a CNN (convolutional neural network). The proposed algorithm consists of a feature extraction unit which extracts the main features of each parallax image and a depth learning unit which learns the parallax information using extracted features. First, the feature extraction unit extracts a feature map for each parallax image through the Xception module and the ASPP(Atrous spatial pyramid pooling) module, which are composed of 2D CNN layers. Then, the feature map for each parallax is accumulated in 3D form according to the time difference and the depth image is estimated after passing through the depth learning unit for learning the depth estimation weight through 3D CNN. The proposed algorithm estimates the depth of object region more accurately than other algorithms.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.2
• /
• pp.714-720
• /
• 2021

The ITL (information theoretic learning) based on the kernel density estimation method that has successfully been applied to machine learning and signal processing applications has a drawback of severe sensitiveness in choosing proper kernel sizes. For the maximization of correntropy criterion (MCC) as one of the ITL-type criteria, several methods of adapting the remaining kernel size ( ) after removing the term have been studied. In this paper, it is shown that the main cause of sensitivity in choosing the kernel size derives from the term and that the adaptive adjustment of in the remaining terms leads to approach the absolute value of error, which prevents the weight adjustment from continuing. Thus, it is proposed that choosing an appropriate constant as the kernel size for the remaining terms is more effective. In addition, the experiment results when compared to the conventional algorithm show that the proposed method enhances learning performance by about 2dB of steady state MSE with the same convergence rate. In an experiment for channel models, the proposed method enhances performance by 4 dB so that the proposed method is more suitable for more complex or inferior conditions.

• Journal of the Architectural Institute of Korea Structure & Construction
• /
• v.36 no.3
• /
• pp.113-120
• /
• 2020

The purpose of this study is to compare the models of Deep Learning-based Convolution Neural Network(CNN) for concrete crack detection. The comparison models are AlexNet, GoogLeNet, VGG16, VGG19, ResNet-18, ResNet-50, ResNet-101, and SqueezeNet which won ImageNet Large Scale Visual Recognition Challenge(ILSVRC). To train, validate and test these models, we constructed 3000 training data and 12000 validation data with 256×256 pixel resolution consisting of cracked and non-cracked images, and constructed 5 test data with 4160×3120 pixel resolution consisting of concrete images with crack. In order to increase the efficiency of the training, transfer learning was performed by taking the weight from the pre-trained network supported by MATLAB. From the trained network, the validation data is classified into crack image and non-crack image, yielding True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), and 6 performance indicators, False Negative Rate (FNR), False Positive Rate (FPR), Error Rate, Recall, Precision, Accuracy were calculated. The test image was scanned twice with a sliding window of 256×256 pixel resolution to classify the cracks, resulting in a crack map. From the comparison of the performance indicators and the crack map, it was concluded that VGG16 and VGG19 were the most suitable for detecting concrete cracks.

• Journal of Korea Multimedia Society
• /
• v.25 no.10
• /
• pp.1375-1385
• /
• 2022

In this paper, we propose a semi-supervised domain adaptation solution to deal with practical face recognition (FR) scenarios where a single face image for each target identity (to be recognized) is only available in the training phase. Main goal of the proposed method is to reduce the discrepancy between the target and the source domain face images, which ultimately improves FR performances. The proposed method is based on the Domain Adatation network (DAN) using an MMD loss function to reduce the discrepancy between domains. In order to train more effectively, we develop a novel loss function learning strategy in which MMD loss and cross-entropy loss functions are adopted by using different weights according to the progress of each epoch during the learning. The proposed weight adoptation focuses on the training of the source domain in the initial learning phase to learn facial feature information such as eyes, nose, and mouth. After the initial learning is completed, the resulting feature information is used to training a deep network using the target domain images. To evaluate the effectiveness of the proposed method, FR performances were evaluated with pretrained model trained only with CASIA-webface (source images) and fine-tuned model trained only with FERET's gallery (target images) under the same FR scenarios. The experimental results showed that the proposed semi-supervised domain adaptation can be improved by 24.78% compared to the pre-trained model and 28.42% compared to the fine-tuned model. In addition, the proposed method outperformed other state-of-the-arts domain adaptation approaches by 9.41%.