• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.631-634
• /
• 2004

In this paper, the Bayesian recurrent neural network (BRNN) is proposed to predict time series data. Among the various traditional prediction methodologies, a neural network method is considered to be more effective in case of non-linear and non-stationary time series data. A neural network predictor requests proper learning strategy to adjust the network weights, and one need to prepare for non-linear and non-stationary evolution of network weights. The Bayesian neural network in this paper estimates not the single set of weights but the probability distributions of weights. In other words, we sets the weight vector as a state vector of state space method, and estimates its probability distributions in accordance with the Bayesian inference. This approach makes it possible to obtain more exact estimation of the weights. Moreover, in the aspect of network architecture, it is known that the recurrent feedback structure is superior to the feedforward structure for the problem of time series prediction. Therefore, the recurrent network with Bayesian inference, what we call BRNN, is expected to show higher performance than the normal neural network. To verify the performance of the proposed method, the time series data are numerically generated and a neural network predictor is applied on it. As a result, BRNN is proved to show better prediction result than common feedforward Bayesian neural network.

• Journal of Korean Academy of Nursing
• /
• v.44 no.6
• /
• pp.708-716
• /
• 2014

Purpose: The purpose of this study was to evaluate the nutritional status of low-income urban elders by diversified ways, and to analyze the risk factors for malnutrition. Methods: The participants in this study were 183 low-income elders registered at a visiting healthcare facility in a public health center. Data were collected using anthropometric measurements, and a questionnaire survey. For data analysis, descriptive statistics, ${\chi}^2$-test, t-test, Fisher's exact test, multiple logistic regression analysis were performed using SPSS 20.0. Results: Regarding the nutritional status of low-income elders as measured by the Mini Nutritional Assessment (MNA), 10.4% of the elders were classified as malnourished; 57.4% as at high risk for malnutrition; and 32.2% as having normal nutrition levels. The main factors affecting malnutrition for low-income elders were loss of appetite (OR=3.34, 95% CI: 1.16~9.56) and difficulties in meal preparation (OR=2.35, 95% CI: 1.13~4.88). Conclusion: In order to effectively improve nutrition in low-income urban elders, it is necessary to develop individual intervention strategies to manage factors that increase the risk of malnutrition and to use systematic approach strategies in local communities in terms of a nutrition support system.

• Journal of Korean Institute of Industrial Engineers
• /
• v.20 no.3
• /
• pp.151-166
• /
• 1994

The manufacturing environment on which this research is focused is an FMS in which AGVs are used for material handling and each part type has one or more process plans. The research aims at developing a methodology whereby, given a part and volume mix for production during any production session, the best set of process plans including one plan per part type is selected and the best unidirectional AGV guidepath can be dynamically reconfigured in response to changes in parts and lot sizes combination. For the integrated PPS/FGD problem in which two functions of process plan selection (PPS) and flexible AGV guidepath design (FGD) are integrated, a zero-one integer programming model is developed. The integrated problem is decomposed into two subproblems, process plan selection given a directed AGV layout and AGV guidepath design with a fixed process plan per part type. A heuristic algorithm that alternately and iteratively solves these two subproblems is developed. The effectiveness of the heuristic algorithm is tested by solving various randomly generated sample problems and comparing the heuristic solutions with those obtained by an exact procedure. From the test results, the following conclusions are drawn: 1) For a reasonable size problem, the heuristic is very effective. 2) By integrating the two functions of PPS and FGD, a remarkable benefit in total production time for a given part and volume mix is gained.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.26 no.2
• /
• pp.406-414
• /
• 2002

A new approach, referred to as a successive zooming genetic algorithm (SZGA), is Proposed for identifying a global solution for continuous optimization problems. In order to improve the local fine-tuning capability of GA, we introduced a new method whereby the search space is zoomed around the design point with the best fitness per 100 generation. Furthermore, the reliability of the optimized solution is determined based on the theory of probability. To demonstrate the superiority of the proposed algorithm, a simple genetic algorithm, micro genetic algorithm, and the proposed algorithm were tested as regards for the minimization of a multiminima function as well as simple functions. The results confirmed that the proposed SZGA significantly improved the ability of the algorithm to identify a precise global minimum. As an example of structural optimization, the SZGA was applied to the optimal location of support points for weight minimization in the radial gate of a dam structure. The proposed algorithm identified a more exact optimum value than the standard genetic algorithms.

• Journal of Information Technology Services
• /
• v.12 no.1
• /
• pp.323-340
• /
• 2013

As the importance of collaboration becomes critical in today's open and complex business environment network, the issues and solutions on compatibility and reusability between different kinds of applications are being increasingly important as well in systems analysis and design. And therefore, service-centered SOA is receiving attention in such business environment as a strategic approach that makes possible for prompt action according to the needs of users and business process. Various implementation methodologies have been proposed for SOA, however, in practical aspects most of them have some problems since they fail to propose specific policies in definition and identification of services for the exact user requirements and business situations. To solve or alleviate those problems, this paper suggests a new service identification model based on hierarchical ontology, where three different ontologies such as business ontology, context ontology and service ontology are proposed to define the relationship and design the link between user requirements, business process, applications and services. Through a suggested methodology in this paper, it would be possible to provide proactive services that meets a variety of business environments and demands of user. Also, since the information can be modified adaptively and dynamically by hierarchical ontology, this study is expected to play a positive role in increasing the flexibility of systems and business environments.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.22 no.8
• /
• pp.742-747
• /
• 2012

A modified NDIF method using a sub-domain approach is introduced to extract highly accurate eigenvalues of two-dimensional, arbitrarily shaped acoustic cavities. The NDIF method, which was developed by the authors for the eigen-mode analysis of arbitrarily shaped acoustic cavities, has the feature that it yields highly accurate eigenvalues compared with other analytical methods or numerical methods(FEM and BEM). However, the NDIF method has the weak point that it can be applicable for only convex cavities. It was revealed that the solution of the NDIF method is very inaccurate or is not suitable for concave cavities. To overcome the weak point, the paper proposes the sub-domain method of dividing a concave domain into several convex domains. Finally, the validity of the proposed method is verified in two case studies, which indicate that eigenvalues obtained by the proposed method are more accurate compared to the exact method, the NDIF method, or FEM(ANSYS).

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.23 no.4
• /
• pp.347-355
• /
• 2013

A damage estimation method for monopile support structure of offshore wind turbine using modal properties and committee of neural networks is presented for effective structural health monitoring. An analytical model for a monopile support structure is established, and the natural frequencies, mode shapes, and mode shape slopes for the support structure are calculated considering soil condition and added mass. The input to the neural networks consists of the modal properties and the output is composed of the stiffness indices of the support structure. Multiple neural networks are constructed and each individual network is trained independently with different initial synaptic weights. Then, the estimated stiffness indices from different neural networks are averaged. Ten damage cases are estimated using the proposed method, and the identified damage locations and severities agree reasonably well with the exact values. The accuracy of the estimation can be improved by applying the committee of neural networks which is a statistical approach averaging the damage indices in the functional space.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.4 no.3
• /
• pp.585-594
• /
• 2000

The exact path following of an autonomous mobile robot in a factory and an unreliable environment has many disadvantages in case of a classical control algorithm. In this paper, a neural network control approach based on an error back propagation algorithm is proposed for controlling a mobile robot to follow a line installed on the road. Since not only the three recognized informations from three sensors attached on a mobile robot but also the ten detailed informations in non recognition area are learned with input patterns, a mobile robot moves smoothly an installed line in spite of non perception space. The mobile robot has an effect of error minimization with a short time till a destination. To test an effectiveness of the proposed controller, the two motor velocity changes which is affected from a moving angle change of a mobile robot are simulated with computer.

• Journal of Korea Society of Industrial Information Systems
• /
• v.6 no.3
• /
• pp.115-122
• /
• 2001

A scene change detection is an important step for video indexing and retrieval. This paper proposes an algorithm by a phased algorithm for fast and accurate detection of abrupt scene changes in an MPEG compressed domain with minimal decoding requirements and computational effort. The proposed method compares two successive I-frames for locating a scene change occurring within the GOP and uses macroblock-coded type information contained in B-frames to detect the exact frame where the scene change occurred. The algorithm has the advantage of speed, simplicity and accuracy. In addition, it requires less amount of storage. The experiment results demonstrate that the proposed algorithm has better detection performance, such as precision and recall rate, than the existing method using all DC images.

• Journal of Korea Multimedia Society
• /
• v.19 no.11
• /
• pp.1837-1851
• /
• 2016

Newspaper reader mobile applications using text-to-speech (TTS) function enable blind people to read newspaper contents. But, tables cannot be easily read by the reader program because most of the tables are stored as images in the contents. Even though we try to use OCR (Optical character reader) programs to recognize letters from the table images, it cannot be simply applied to the table reading function because the table structure is unknown to the readers. Therefore, identification of exact location of each table cell that contains the text of the table is required beforehand. In this paper, we propose an efficient image processing algorithm to recognize all the cells in tables by identifying columns and rows in table images. From the cell location data provided by the table column and row identification algorithm, we can generate table structure information and table reading scenarios. Our experimental results with table images found commonly in newspapers show that our cell identification approach has 100% accuracy for simple black and white table images and about 99.7% accuracy for colored and complicated tables.