• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.4
• /
• pp.416-426
• /
• 2003

In this paper we propose a new method for choosing the initial values of Expectation-Maximization(EM) algorithm that has been used in various applications for clustering. Conventionally, the initial values were chosen randomly, which sometimes yields undesired local convergence. Later, K-means clustering method was employed to choose better initial values, which is currently widely used. However the method using K-means still has the same problem of converging to local points. In order to resolve this problem, a new method of initializing values for the EM process. The proposed method not only strengthens the characteristics of EM such that the number of iteration is reduced in great amount but also removes the possibility of falling into local convergence.

• Journal of Environmental Science International
• /
• v.26 no.2
• /
• pp.249-256
• /
• 2017

In this study, we evaluated the potential of 70% ethanol extract from Persicaria nepalensis (PNE) as a cosmetic ingredient by primary skin irritation, ocular irritation, and maximization tests for delayed hypersensitivity in New Zealand white rabbits and Hartley guinea pig. Skin safety study was performed to evaluate the potential toxicity of PNE using the primary irritation test. In the primary irritation test, 50% PNE was applied to the skin, and no adverse reactions such as erythema and edema were observed at the intact skin sites. Therefore, PNE was classified as a practically non-irritating material based on a primary irritation index of "0.0.". In the ocular irritation test, the 50% PNE applied did not show any adverse reactions in the different parts of rabbit eyes, including the cornea, iris, and conjunctiva. Thus, PNE was classified as a practically non-irritating material based on an acute ocular irritation index of "0.0.". Skin sensitization was tested by the Guinea Pig Maximization Test (GPMT) and Freund's Complete Adjuvant (FCA) using an intradermal injection of 10% PNE. Edema and erythema were not observed 24 and 48 h after the topical application of PNE in skin sensitization test, which exhibited a sensitization score of "0.0.". Therefore, it can be suggested that P. nepalensis could be used as potential candidates for cosmoceutical ingredients, without any major side effects.

• 전자공학회논문지 IE
• /
• v.45 no.3
• /
• pp.20-26
• /
• 2008

Two independent component analysis(ICA) algorithms were applied for the recognition of speech signals corrupted by a car engine noise. Speech recognition was performed by hidden markov model(HMM) for the estimated signals and recognition rates were compared with those of orginal speech signals which are not corrupted. Two different ICA methods were applied for the estimation of speech signals, one of which is FastICA algorithm that maximizes negentropy, the other is information-maximization approach that maximizes the mutual information between inputs and outputs to give maximum independence among outputs. Word recognition rate for the Korean news sentences spoken by a male anchor is 87.85%, while there is 1.65% drop of performance on the average for the estimated speech signals by FastICA and 2.02% by information-maximization for the various signal to noise ratio(SNR). There is little difference between the methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.9
• /
• pp.4357-4378
• /
• 2017

In cooperative multi-relay networks, the relay nodes which are selected are very important to the system performance. How to choose the best cooperative relay nodes is an optimization problem. In this paper, multi-relay selection schemes which consider either single objective or multi-objective are proposed based on evolutionary algorithms. Firstly, the single objective optimization problems of multi-relay selection considering signal to noise ratio (SNR) or power efficiency maximization are solved based on the quantum bee colony optimization (QBCO). Then the multi-objective optimization problems of multi-relay selection considering SNR maximization and power consumption minimization (two contradictive objectives) or SNR maximization and power efficiency maximization (also two contradictive objectives) are solved based on non-dominated sorting quantum bee colony optimization (NSQBCO), which can obtain the Pareto front solutions considering two contradictive objectives simultaneously. Simulation results show that QBCO based multi-relay selection schemes have the ability to search global optimal solution compared with other multi-relay selection schemes in literature, while NSQBCO based multi-relay selection schemes can obtain the same Pareto front solutions as exhaustive search when the number of relays is not very large. When the number of relays is very large, exhaustive search cannot be used due to complexity but NSQBCO based multi-relay selection schemes can still be used to solve the problems. All simulation results demonstrate the effectiveness of the proposed schemes.

• The Journal of the Korea Contents Association
• /
• v.17 no.5
• /
• pp.285-304
• /
• 2017

This paper introduces household's decision for schooling and firm's decision for technology-specific training together into the second generation model of endogenous innovation, and analyses how schooling and technology-specific training interact each other, how they respectively affect innovation and economic growth, and also how the portfolio mix of schooling and technology-specific training changes as economy becomes more innovative. Main results are as follows: First, schooling and technology-specific training both have "inverted-U"shape growth effects. Second, schooling investment per labor required for growth maximization is always greater than that for firm profit maximization. Third, the optimal schooling for growth maximization decreases with technology-specific training. Fourth, the schooling effect on technology-specific training is "U"shaped, so that for firm's profit maximization schooling is substitutable for technology-specific training at the relatively lower level of schooling but complementary at its relatively higher level. Fifth, as economy becomes more innovative, the portfolio mix of education changes in favor of schooling.

• Journal of KIISE:Software and Applications
• /
• v.30 no.5_6
• /
• pp.454-460
• /
• 2003

In this paper, we propose an efficient method for recognizing human lips. Based on Point Distribution Model, a lip shape is represented as a set of points. We calculate a lip model and the distribution of shape parameters using Principle Component Analysis and Gaussian mixture, respectively. The Expectation Maximization algorithm is used to determine the maximum likelihood parameter of Gaussian mixture. The lip contour model is derived by using the gray value changes at each point and in regions around the point and used to search the lip shape in a image. The experiments have been performed for many images, and show very encouraging result.

• The Korean Journal of Applied Statistics
• /
• v.28 no.3
• /
• pp.467-475
• /
• 2015

Accurate demand forecasting is a crucial component in revenue management(RM). The booking data of departed flights is used to forecast the demand for future departing flights; however, some booking requests that were denied were omitted in the departed flights data. Denied booking requests can be interpreted as censored in statistics. Thus, unconstraining demand is an important issue to forecast the true demands of future flights. Several unconstraining methods have been introduced and a method based on expectation maximization is considered superior. In this study, we propose a new unconstraining method based on a regression model that can entertain such censored data. Through a simulation study, the performance of the proposed method was evaluated with two representative unconstraining methods widely used in RM.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2006.11a
• /
• pp.938-941
• /
• 2006

We present a new design method of one-dimensional multi-layered acoustic foams for transmission loss maximization by topology optimization. Multi-layered acoustic foam sequences consisting of acoustic air layers and poroelastic material layers are designed for target frequency values. For successful topology optimization design of multi-layered acoustic foams, the material interpolation concept of topology optimization is adopted. In doing so, an acoustic air layer is modeled as a limiting poroelastic material layer; acoustic air and poroelastic material are handled by a single set of governing equations based on Biot's theory. For efficient analysis of a specific multi-layered foam appearing during optimization, we do not solve the differential equations directly, but we use an efficient transfer matrix approach which can be derived from Biot's theory. Through some numerical case studies, the proposed design method for finding optimal multi-layer sequencing is validated.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.16 no.5
• /
• pp.104-111
• /
• 2002

This paper presents a fuzzy rule extraction method using EM(Expectation-Maximization) algorithm and a design method of adaptive neuro-fuzzy control. EM algorithm is used to estimate a maximum likelihood of a GMM(Gaussian Mixture Model) and cluster centers. The estimated clusters is used to automatically construct the fuzzy rules and membership functions for ANFIS(Adaptive Neuro-Fuzzy Inference System). Finally, we applied the proposed method to the water temperature control system and obtained better results with respect to the number of rules and SAE(Sum of Absolute Error) than previous techniques such as conventional fuzzy controller.

• Journal of KIISE
• /
• v.44 no.10
• /
• pp.1105-1111
• /
• 2017

Influence maximization (IM) is the problem of finding a seed set composed of k nodes that maximizes the influence spread in social networks. However, one of the biggest problems of existing solutions for IM is that it takes too much time to select a k-seed set. This performance issue occurs at the micro and macro levels. In this paper, we propose a fast hybrid method that addresses two issues at micro and macro levels. Furthermore, we propose a path-based community detection method that helps to select a good seed set. The results of our experiment with four real-world datasets show that the proposed method resolves the two issues at the micro and macro levels and selects a good k-seed set.