• Journal of Semiconductor Engineering
• /
• v.1 no.1
• /
• pp.31-37
• /
• 2020

We present an approximate DRAM architecture for energy-efficient deep learning. Our key premise is that by bounding memory errors to non-critical information, we can significantly reduce DRAM refresh energy without compromising recognition accuracy of deep neural networks. To validate the key premise, we make extensive Monte-Carlo simulations for several well-known convolutional neural networks such as LeNet, ConvNet and AlexNet with the input of MINIST, CIFAR-10, and ImageNet, respectively. We assume that the highest-order 8-bits (in single precision) and 4-bits (in half precision) are protected from retention errors under the proposed architecture and then, randomly inject bit-errors to unprotected bits with various bit-error-rates. Here, recognition accuracies of the above convolutional neural networks are successfully maintained up to the 10^-5-order bit-error-rate. We simulate DRAM energy during inference of the above convolutional neural networks, where the proposed architecture shows the possibility of considerable energy saving up to 10 ~ 37.5% of total DRAM energy.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2023.05a
• /
• pp.179-180
• /
• 2023

In this study, the performance of the prediction model was compared and analyzed using DNN and LSTM learning models to predict the amount of dry shrinkage of the concrete. As a result of the analysis, DNN model had a high error rate of about 51%, indicating overfitting to the training data. But, the LSTM learning model showed a relatively higher accuracy with an error rate of 12% compared to the DNN model. Also, the Pre_LSTM model which preprocess data, showed the performance with an error rate of 9% and a coefficient of determination of 0.887 in the LSTM learning model.

• Journal of Auto-vehicle Safety Association
• /
• v.13 no.4
• /
• pp.129-143
• /
• 2021

This paper proposes actuator fault detection and adaptive fault-tolerant control algorithms using performance index and human-like learning for longitudinal autonomous vehicles. Conventional longitudinal controller for autonomous driving consists of supervisory, upper level and lower level controllers. In this paper, feedback control law and PID control algorithm have been used for upper level and lower level controllers, respectively. For actuator fault-tolerant control, adaptive rule has been designed using the gradient descent method with estimated coefficients. In order to adjust the control parameter used for determination of adaptation gain, human-like learning algorithm has been designed based on perceptron learning method using control errors and control parameter. It is designed that the learning algorithm determines current control parameter by saving it in memory and updating based on the cost function-based gradient descent method. Based on the updated control parameter, the longitudinal acceleration has been computed adaptively using feedback law for actuator fault-tolerant control. The finite window-based performance index has been designed for detection and evaluation of actuator performance degradation using control error.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.3868-3888
• /
• 2022

A widely used social networking service like Twitter has the ability to disseminate information to large groups of people even during a pandemic. At the same time, it is a convenient medium to share irrelevant and unverified information online and poses a potential threat to society. In this research, conventional machine learning algorithms are analyzed to classify the data as either non-rumor data or rumor data. Machine learning techniques have limited tuning capability and make decisions based on their learning. To tackle this problem the authors propose a deep learning-based Rumor Detection Neural Network model to predict the rumor tweet in real-world events. This model comprises three layers, AttCNN layer is used to extract local and position invariant features from the data, AttBi-LSTM layer to extract important semantic or contextual information and HPOOL to combine the down sampling patches of the input feature maps from the average and maximum pooling layers. A dataset from Kaggle and ground dataset #gaja are used to train the proposed Rumor Detection Neural Network to determine the veracity of the rumor. The experimental results of the RDNN Classifier demonstrate an accuracy of 93.24% and 95.41% in identifying rumor tweets in real-time events.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.427-428
• /
• 2021

정보통신기술의 발달로 전자상거래의 증가와 소비자들의 제품에 대한 경험과 지식의 공유가 활발하게 진행됨에 따라 소비자는 제품을 구매하기 위한 자료수집, 활용을 진행하고 있다. 따라서 기업은 다양한 기능들을 반영한 제품이 치열하게 경쟁하고 있는 현 시장에서 우위를 점하고자 소비자 리뷰를 분석하여 소비자의 정확한 소비자의 요구사항을 분석하여 제품기획 프로세스에 반영하고자 텍스트마이닝(Text Mining) 기술과 딥러닝(Deep Learning) 기술을 통한 연구가 이루어지고 있다. 본 논문의 기초자료가 되는 데이터셋은 포털사이트의 구매사이트와 오픈마켓 사이트의 소비자 리뷰를 웹크롤링하고 자연어처리하여 진행한다. 감성분석은 딥러닝기술 중 CNN(Convolutional Neural Network), LSTM(Long Short Term Memory) 조합의 모델을 구현한다. 이는 딥러닝을 이용한 제품기획 프로세스로 소비자 요구사항 반영, 경제적인 측면, 제품기획 시간단축 등 긍정적인 영향을 미칠 것으로 기대한다.

• English Language & Literature Teaching
• /
• v.11 no.3
• /
• pp.89-112
• /
• 2005

This study investigated the effects of collocation-based vocabulary instruction for the experimental group (G2). It was compared to the traditional wordlist-based vocabulary instruction for the control group (G1). This results reflect the development of low level high school EFL learners' vocabulary learning strategy use and the positive change in the affective domain. In the analysis of the survey responses, G1 and G2 did not differ significantly on the first questionnaire. They did, however, differ significantly on the second questionnaire. G2 used more strategies to discover and to consolidate the meaning of the words by means of combining words. In terms of the affective domain, G2 participated more actively in the learning activities, which had a significant effect on vocabulary growth, memory, self-confidence, motivation, and cooperative learning. This is attributable to the fact that G2 was more inquisitive, interested, challenged, participatory, cooperative, and attentive than G1 in performing the vocabulary task activities. Moreover, the data collected from the questionnaire showed that G2 performed more interactive and dynamic activities in solving the given tasks.

• Journal of the Korean Society for Heat Treatment
• /
• v.36 no.6
• /
• pp.374-381
• /
• 2023

This study develops and evaluates a deep learning model for predicting oxide and nitride layers based on plasma process data. We introduce a novel deep learning-based Varying Coefficient Regressor (VCR) by adapting the VCR, which previously relied on an existing unique function. This model is employed to forecast the oxide and nitride layers within the plasma. Through comparative experiments, the proposed VCR-based model exhibits superior performance compared to Long Short-Term Memory, Random Forest, and other methods, showcasing its excellence in predicting time series data. This study indicates the potential for advancing prediction models through deep learning in the domain of plasma processing and highlights its application prospects in industrial settings.

• The Journal of Korean Medicine
• /
• v.30 no.3
• /
• pp.1-14
• /
• 2009

Objectives : Alzheimer's disease (AD) is characterized by neuronal loss and extracellular senile plaque. Moreover, the cellular actions of ${\beta}$-amyloid (A${\beta}$ play a causative role in the pathogenesis of AD. This study was designed to determine whether Woo-Gui-Um, a commonly used Korean herbal medicine, has the ability to protect cortical and hippocampal neurons against A${\beta}_{25-35}$ neurotoxicity Methods : In the present study, the authors investigated the preventative effects of the water extract of Woo-Gui-Um in a mouse model of AD. Memory impairment was induced by intraventricularly (i.c.v.) injecting A${\beta}_{25-35}$ peptides into mice. Woo-Gui-Um extract was then administered orally (p.o.) for 14 days. In addition, A${\beta}_{25-35}$ toxicity on the hippocampus was assessed immunohistochemically, by staining for Tau, MAP2, TUNEL, and Bax, and by performing an in vitro study in PC12 cells. Results : Woo-Gui-Um extract had an effect to improve learning ability and memory score in the water maze task. Woo-Gui-Um extract had significant neuroprotective effects in vivo against oxidative damage and apoptotic cell death of hippocampal neurons caused by i.c.v. A${\beta}_{25-35}$. In addition, Woo-Gui-Um extract was found to have a protective effect on A${\beta}_{25-35}$-induced apoptosis, and to promote neurite outgrowth of nerve growth factor (NGF)-differentiated PC12 cells. Conclusions : These results suggest that Woo-Gui-Um extract reduces memory impairment and Alzheimer's dementia via an anti-apoptotic effect and by regulating Tau and MAP2 in the hippocampus.

• Cross-Cultural Studies
• /
• v.39
• /
• pp.407-432
• /
• 2015

This study focuses on the flow of story and content or related context when using movies as learning materials in a class. A great advantage of using movies is that they have a consistent story and detailed content development. Most teachers, however, tend to concentrate on practical expressions totally unrelated to the story or context of the movie they are using. This way might be efficient in the short run but it is certain that the expressions are unlikely to be retained in long-term memory. This study examines how a story-centered class influences learning of practical expressions and how efficient this approach to learning is. Learning and teaching with focus only on the expressions in a movie shades the meaning of the use of the movie a little. In this study the movie, Cars 2, was used in a course of general education with 150 students enrolled. Various group activities were suggested to immerse students into the story and contents of Cars 2. It was found that a story-centered class is helpful for students to acquire practical expressions and that students' satisfaction level with the class was high.

• 대한공업교육학회지
• /
• v.37 no.2
• /
• pp.147-160
• /
• 2012

The purpose of this study was to examine the effects on the academic achievement and learning satisfaction according to the presentation method of the multimedia materials for 'transportation technology' units of technology home economics subject. The subjects were assigned in third conditions; Text type explanation class, multimedia class and multimedia video class with narration. The data of six evaluation questions obtained from the survey of 93 high school girl were analyzed using SPSS program. The results of the study were as follows : First, in the learning satisfaction average level(M) of the students' overall responses to the questions, multimedia teaching learning class(experimental group 1) is the first(M=4.14), multimedia video class with narration(experimental group 2) is the second(M=3.16), and instructor-led class(control group) is the third (M=2.63). Therefore, the teaching learning multimedia class(experimental group 1) was most effective. Second, looking at the correlations between the students' responses to the questions, in an interesting class, the students have a retentive memory and comprehension, but a lower concentration can not a retentive memory. Third, multimedia teaching learning class(experimental group 1) has the best degree at the level of academic achievement, but instructor-led class(control group) and multimedia video class with narration(experimental group 2) have similar degree in the second place. To increase academic achievement, an instructor-led class is important to arouse interest and a multimedia video class with narration is required ways to improve level of concentration.