• International Journal of Control, Automation, and Systems
• /
• v.1 no.1
• /
• pp.101-110
• /
• 2003

This paper introduces an identification method for nonlinear models in the form of rule-based Fuzzy-Neural Networks (FNN). In this study, the development of the rule-based fuzzy neural networks focuses on the technologies of Computational Intelligence (CI), namely fuzzy sets, neural networks, and genetic algorithms. The FNN modeling and identification environment realizes parameter identification through synergistic usage of clustering techniques, genetic optimization and a complex search method. We use a HCM (Hard C-Means) clustering algorithm to determine initial apexes of the membership functions of the information granules used in this fuzzy model. The parameters such as apexes of membership functions, learning rates, and momentum coefficients are then adjusted using the identification algorithm of a GA hybrid scheme. The proposed GA hybrid scheme effectively combines the GA with the improved com-plex method to guarantee both global optimization and local convergence. An aggregate objective function (performance index) with a weighting factor is introduced to achieve a sound balance between approximation and generalization of the model. According to the selection and adjustment of the weighting factor of this objective function, we reveal how to design a model having sound approximation and generalization abilities. The proposed model is experimented with using several time series data (gas furnace, sewage treatment process, and NOx emission process data from gas turbine power plants).

• Transactions on Control, Automation and Systems Engineering
• /
• v.3 no.1
• /
• pp.43-50
• /
• 2001

In this paper, we propose the design procedure of advance Polynomial Neural Networks(PNN) architecture for optimal model identification of complex and nonlinear system. The proposed PNN architecture is presented as the generic and advanced type. The essence of the design procedure dwells on the Group Method of Data Handling(GMDH). PNN is a flexible neural architecture whose structure is developed through learning. In particular, the number of layers of the PNN is not fixed in advance but is generated in a dynamic way. In this sense, PNN is a self-organizing network. With the aid of three representative numerical examples, compari-sons show that the proposed advanced PNN algorithm can produce the model with higher accuracy than previous other works. And performance index related to approximation and generalization capabilities of model is evaluated and also discussed.

• Journal of KIISE:Software and Applications
• /
• v.27 no.2
• /
• pp.97-107
• /
• 2000

Incremental learning generally reduces training time and increases the generalization of a neural network by selecting training data incrementally during the training. However, the existing methods of incremental learning repeatedly evaluate the importance of training data every time they select additional data. In this paper, an incremental learning algorithm is proposed for pattern classification problems. It evaluates the importance of each piece of data only once before starting the training. The importance of the data depends on how close they are to the decision boundary. The current paper presents an algorithm which orders the data according to their distance to the decision boundary by using clustering. Experimental results of two artificial and real world classification problems show that this proposed incremental learning method significantly reduces the size of the training set without decreasing generalization performance.

• Journal of Biomedical Engineering Research
• /
• v.42 no.4
• /
• pp.159-166
• /
• 2021

In order to analyze cell images, accurate segmentation of each cell is indispensable. However, the reality is that accurate cell image segmentation is not easy due to various noises, dense cells, and inconsistent shape of cells. Therefore, in this paper, we propose an algorithm that combines marker-based watershed segmentation and ellipse fitting method for glioblastoma cell segmentation. In the proposed algorithm, in order to solve the over-segmentation problem of the existing watershed method, the marker-based watershed technique is primarily performed through "seeding using local minima". In addition, as a second process, the concave point search using ellipse fitting for final segmentation based on the connection line between the concave points has been performed. To evaluate the performance of the proposed algorithm, we compared three algorithms with other algorithms along with the calculation of segmentation accuracy, and we applied the algorithm to other cell image data to check the generalization and propose a solution.

• Journal of the Korea Management Engineers Society
• /
• v.23 no.4
• /
• pp.125-141
• /
• 2018

As global competition is intensifying, companies are entering not only developed markets but also emerging and developing countries. In emerging and developing countries, business-to-governmental relations are an essential factor influencing the firm performance. The purpose of this research is to examine the effects of corporate political activities on firm performance. Using the BEEPS data developed by the EBRD and the World Bank, the effect of various bribery activities regarded as corporate political activities on the firm performance was examined. Also, the moderating effect of competition type on the firm performance was proved. As the results, the bribery contribution, which is a corporate political activity, has a positive effect on the firm performance. Specifically, when infrastructure bribery occurred, costs and sales were higher, and when permission bribery was granted, profits, costs, and sales were higher. In addition, the effect of bribery as a corporate political activity on firm performance is moderated by competition type. These findings have academic and practical implications for empirical analyzing of SMEs' firm performance, guidelines suggestion for entry into the other countries, generalization possibility of using credible data, developing and designing export support systems and programs. Finally, limitations of the research and future research directions were discussed.

• International Journal of Contents
• /
• v.10 no.4
• /
• pp.69-74
• /
• 2014

General location aware systems are only applied to indoor and outdoor environments using more than three transmitters to estimate a fixed object location. Those kinds of systems have environmental restrictions that require an already established infrastructure. To solve this problem, an Object Location Estimation (OLE) algorithm based on PTP (Point To Point) communication has been proposed. However, the problem with this method is that deduction of performance parameters is not enough and location estimation is very difficult because of unknown restriction conditions. From experimental tests in this research, we determined that the performance parameters for restriction conditions are a maximum transmission distance of CSS communication and an optimum moving distance interval between personal locations. In this paper, a system applied OLE algorithm based on PTP communication is implemented using a CSS (Chirp Spread Spectrum) communication module. A maximum transmission distance for CSS communication and an optimum moving distance interval between personal locations are then deducted and studied to estimate a fixed object location for generalization.

• Electronics and Telecommunications Trends
• /
• v.35 no.3
• /
• pp.9-19
• /
• 2020

Recently, a technique for applying a deep learning language model pretrained from a large corpus to fine-tuning for each application task has been widely used as a language processing technology. The pretrained language model shows higher performance and satisfactory generalization performance than existing methods. This paper introduces the major research trends related to deep learning pretrained language models in the field of language processing. We describe in detail the motivations, models, learning methods, and results of the BERT language model that had significant influence on subsequent studies. Subsequently, we introduce the results of language model studies after BERT, focusing on SpanBERT, RoBERTa, ALBERT, BART, and ELECTRA. Finally, we introduce the KorBERT pretrained language model, which shows satisfactory performance in Korean language. In addition, we introduce techniques on how to apply the pretrained language model to Korean (agglutinative) language, which consists of a combination of content and functional morphemes, unlike English (refractive) language whose endings change depending on the application.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2004.04a
• /
• pp.241-246
• /
• 2004

This paper proposed trajectory tracking control of mobile robot. Trajectory tracking control scheme are real coding genetic-algorithm and back-propergation algorithm. Control scheme ability experience proposed simulation. Stable tracking control problem of mobile robots have been studied in recent years. These studios have guaranteed stability of controller, but the performance of transient state has not been guaranteed. In some situations, constant gain controller shows overshoots and oscillations. So we introduce better control scheme using Real coding Genetic Algorithm(RCGA) and neural network. Using RCGA, we can find proper gains in several situations and these gains are generalized by neural network. The generalization power of neural network will give proper gain in untrained situation. Performance of proposed controller will verify numerical simulations and the results show better performance than constant gain controller.

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.13 no.3
• /
• pp.126-132
• /
• 2004

This paper proposed trajectory tracking control based on genetic algorithm. Trajectory tracking control scheme are real coding genetic algorithm(RCGA) and back-propagation algorithm(BPA). Control scheme ability experience proposed simulation. Stable tracking control problem of mobile robots have been studied in recent years. These studies have guaranteed stability of controller, but the performance of transient state has not been guaranteed. In some situations, constant gain controller shows overshoots and oscillations. So we introduce better control scheme using real coding genetic algorithm and neural network. Using RCGA, we can find proper gains in several situations and these gains are generalized by neural network. The generalization power of neural network will give proper gain in untrained situation. Performance of proposed controller will verity numerical simulations and the results show better performance than constant gain controller.

• Communications for Statistical Applications and Methods
• /
• v.13 no.1
• /
• pp.151-165
• /
• 2006

It is important to estimate the true misclassification rate of a given classifier when an independent set of test data is not available. Cross-validation and bootstrap are two possible approaches in this case. In related literature bootstrap estimators of the true misclassification rate were asserted to have better performance for small samples than cross-validation estimators. We compare the two estimators empirically when the classification rule is so adaptive to training data that its apparent misclassification rate is close to zero. We confirm that bootstrap estimators have better performance for small samples because of small variance, and we have found a new fact that their bias tends to be significant even for moderate to large samples, in which case cross-validation estimators have better performance with less computation.