• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.17 no.4
• /
• pp.54-62
• /
• 2018

As a part of realization of artificial intelligence signal(AI Signal), this study proposed an actuated signal algorithm based on vehicle queue length that estimates in real time by deep learning. In order to implement the algorithm, we built an API(COM Interface) to control the micro traffic simulator Vissim in the tensorflow that implements the deep learning model. In Vissim, when the link travel time and the traffic volume collected by signal cycle are transferred to the tensorflow, the vehicle queue length is estimated by the deep learning model. The signal time is calculated based on the vehicle queue length, and the simulation is performed by adjusting the signaling inside Vissim. The algorithm developed in this study is analyzed that the vehicle delay is reduced by about 5% compared to the current TOD mode. It is applied to only one intersection in the network and its effect is limited. Future study is proposed to expand the space such as corridor control or network control using this algorithm.

• Journal of The Korean Association For Science Education
• /
• v.33 no.4
• /
• pp.794-806
• /
• 2013

The purpose of this study is to analyze what kinds of science education theories are targeted in the "Examination for Appointing Secondary School Science Teachers (EASST)." For the analyses, we extracted the contents related to the science education theories in the question items of the EASST of 2008 through 2012, and categorized those theories into science curriculum, history of science and philosophy of science, scientific inquiry, theory of teaching and learning, model of teaching and learning, and assessment. The results of this study indicated that the theory of teaching and learning appeared most frequently and there were high proportions of question items related to the following topics: contents in science curriculum, scientific method, contemporary philosophy of science, process of inquiry, Ausubel's theory, learning cycle model by Lawson, cooperative learning, criteria of performance assessment, and etc. While we, as science educators, believed that the other categories such as 'history of science' provides important topics for pre-service science teachers, questions items dealing with those were rarely found in the past EASSTs. As EASST has strong influences on the professional developments of pre-service science teachers, more research should be pursued on how much and what domains of science education theories would be appropriate for the test.

• Tunnel and Underground Space
• /
• v.32 no.6
• /
• pp.502-517
• /
• 2022

As the use of TBM increases, research has recently increased to to analyze TBM data with machine learning techniques to predict the exchange cycle of disc cutters, and predict the advance rate of TBM. In this study, a regression prediction of disc cutte wear of slurry shield TBM site was made by combining machine learning based on the machine data and the geotechnical data obtained during the excavation. The data were divided into 7:3 for training and testing the prediction of disc cutter wear, and the hyper-parameters are optimized by cross-validated grid-search over a parameter grid. As a result, gradient boosting based on the ensemble model showed good performance with a determination coefficient of 0.852 and a root-mean-square-error of 3.111 and especially excellent results in fit times along with learning performance. Based on the results, it is judged that the suitability of the prediction model using data including mechanical data and geotechnical information is high. In addition, research is needed to increase the diversity of ground conditions and the amount of disc cutter data.

• Journal of Gifted/Talented Education
• /
• v.9 no.2
• /
• pp.73-101
• /
• 1999

The purpose of this study was to suggest an advanced system for educating scientifically gifted children in the Center for Science Talented Education at Kyungpook National University. Several suggestions based on analysis of current identifying-process and instructional materials for scientifically gifted children were provided for advancing the educational system of the center. First, this study suggested a three-step procedure to identify procedure emphasized students reasoning skills as one of important characteristics of the gifted child. Second, this study provided an instructional model for developing hypothesis testing skills in scientifically gifted children. The model was originally based on Lawson's scientific reasoning processes and learning cycle mode. Third, this study also suggested an effective administration system of the Center for Science Talented Education. Further, this study suggested effective ways on research works for advancing the center, educating instructors, the cyber center for remote education, and international co-works for developing the gifted children's potential abilities.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2000.04a
• /
• pp.330-336
• /
• 2000

This paper discusses the potential of system dynamics modelling to support balanced scorecard. The balanced scorecard is a conceptual framework for translating an organisation's strategy into a set of performance indicators. These performance indicators are distributed across the 'classic'model's four perspective: Customers, Internal Business Processes, Financial, and Learning and Growth. This balanced scorecard, whilst having significant strength, suffers from the limitation of all performance indicator systems, namely that the interrelationships between indicators are overlooked and there is no way of taking into account the impact of delayed feedback which flows from introduction of new policy and legislative changes. System Dynamics is a methodology for understanding complex problems where there is dynamic behaviour and where feedback impacts significantly on system outcomes. System dynamics provides a rigorous basis for qualitative testing of the effects of performance indicators in complex environments such as health or social security. This can be supplemented with quantitative system dynamics simulation tools that further test the validity of indicators and the business rules implicit in them. System dynamics modelling has an important role to play in extending feedback cycle in performance measurements to a full systems approach.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.13 no.22
• /
• pp.71-78
• /
• 1990

Some problems to assess the safety of automated man-machine system are studied in many ways. The difficulty occurred in this system is the vagueness of human behavior. Fuzzy set theory is used to assess the human behavior in safety analysis. The unsafe behavior listed top 10 in accident statistics would be explained as the factors of human vagueness. Three cases are considered, which consist of man-machine system as man-man, man-machine, machine-machine types. For the design of safe task, using characteristics of work performance, each motion cycle time is required to know the rate of learning. Approach of human behavior to the standard motion means more safe motion. It is important to design the works as to minimize the time performance to the standard motion's, which utilize the control of risk potential with easy. In that process, use of fuzzy set theory is appropriate to analyze the human behavior to identify its vagueness.

• Proceedings of the Korean Information Science Society Conference
• /
• 2006.06a
• /
• pp.40-42
• /
• 2006

유전자 발현에 있어서의 조절작용은 유전자간의 복합적인 상호작용의 결과에 기인한다. 따라서 이러한 현상으로부터 기능적으로 연관된 유전자 군을 식별하기 위해서는 단일 유전자보다는 복수의 유전자군의 발현패턴을 대상으로 하게 된다. 이 경우 발현패턴의 시간에 따른 다양하고 복잡한 특징들은 은닉변수학습 모형을 이용하므로서 보다 명확하게 표현될 수 있고, 유사한 기능을 가진 유전자 군을 탐색 하는데에 효과적으로 이용될 수 있다. 본 논문에서 제시된 은닉변수학습 모형은 이스트 Cell Cycle 데이터에 적용한 결과 특정 조절유전자에 대하여 생물학적으로 연관된 유전자 군을 찾는 데에 다른 방법과 비교하여 효과적임을 보일 수 있었다.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.8 no.4
• /
• pp.73-81
• /
• 2012

Web development and Internet technology development bring many kinds of works to web. This is the main reason why XML, document standard is popular. XML in web can be used to express document template or standard. XML with java can be more powerful and general. For example, XML can be used to transmit data and to print data into the screen using Ajax in JSP(Java Server Page) and to make interfaces in android, which is useful to reduce development cycle. However, XML is not easy to learn for the novice. In this paper, we propose the easy and effective way to reduce the learning curve of XML and to make and use XML documents. For the purpose, we suggest template base XML document generation and we design and implement XML document generator based on Template. XML document generator of template-based provides user interface and layout of XML document. So, users can generate XML document easily and effectively.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2014.10a
• /
• pp.84-89
• /
• 2014

Data-driven method for fault diagnostics system often use machine learning technique. To use such technique proper signal processing should be implemented such as time synchronous averaging (TSA) for ball bearing systems. However, for journal bearing diagnostics systems not much has been researched, and yet a proper signal processing method has not been studied. Therefore, in this research an optimal datum unit for a reliable journal bearing diagnostics system along with angular resampling process is being suggested. Before extracting time and frequency domain features, angular resampling is applied to each cycle of vibration data. As to preserve the characteristics of vibration signal, averaging method is replaced by finding the optimal datum unit which strengthens statistical characteristics of vibration signal. Then 20 features were extracted for various cases, and those features are being evaluated by two criteria, separability and classification accuracy.