• The KIPS Transactions:PartB
• /
• v.15B no.4
• /
• pp.323-330
• /
• 2008

Since complex diseases are caused by interactions of multiple genes, traditional statistical methods are limited in its power to predict the onset of a complex disease. Recently new approaches using machine learning techniques are introduced. Neural nets are a suitable model to find patterns in complex data. When large amount of data are fed into a neural net, however, it takes a long time for learning and finding patterns. In this study we suggest a new model that combines the set association, which is a statistical technique to find important SNPs associated with complex diseases, and neural network. We experiment with SNP data related to asthma to test the effectiveness of our model. Our model shows higher prediction accuracy and shorter execution time than neural net only. We expect our model can be used effectively to predict the onset of other complex diseases.

• Journal of Internet Computing and Services
• /
• v.5 no.2
• /
• pp.69-73
• /
• 2004

A speaker model In speaker recognition system is to be trained from a large data set gathered in multiple sessions. Large data set requires large amount of memory and computation, and moreover it's practically hard to make users utter the data inseveral sessions. Recently the incremental adaptation methods are proposed to cover the problems, However, the data set gathered from multiple sessions is vulnerable to the outliers from the irregular utterance variations and the presence of noise, which result in inaccurate speaker model. In this paper, we propose an incremental robust adaptation method to minimize the influence of outliers on Gaussian Mixture Madel based speaker model. The robust adaptation is obtained from an incremental version of M-estimation. Speaker model is initially trained from small amount of data and it is adapted recursively with the data available in each session, Experimental results from the data set gathered over seven months show that the proposed method is robust against outliers.

• Asia Marketing Journal
• /
• v.8 no.1
• /
• pp.1-17
• /
• 2006

A couple of previous studies have warned against the use of homogeneous choice models in assessing the effect of reference price since unaccounted for response heterogeneity may result in spurious reference price effects(Chang, Siddarth and Weinberg 1999; Bell and Lattin 2000). According to Meyer and Kahn(1991), not accounting for consideration set heterogeneity may also bias the effect parameters in the choice model. Therefore, failure to account for these two sources of bias, in fact, have cast doubt on the empirical support for reference price effects in general. In view of aforementioned potential sources of bias, the author investigates the robustness of loss aversion effect in the reference-dependent model after accounting for heterogeneity in response as well as consideration set. The proposed model defines individual household's consideration set based on the posterior distribution of preference obtained from the Empirical Bayes approach. In addition, the same posterior distribution is used to form household-specific reference prices. Response heterogeneity correction is carried out via the Latent Class approach. The proposed model outperforms the Reference-Dependent model that includes the reference price measure most often employed in the previous studies. This implies that as a way of simplifying decision task, consumers restrict their consideration set to a subset of available brands not only in making a brand choice but also in forming reference prices.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.28 no.1
• /
• pp.125-132
• /
• 2010

With recent advancement of remote sensing technology, a variety of data acquisition over the same area is achievable. An automated co-registration of heterogeneous airborne images is a critical step for change detection. This paper describes an automatic method for co-registration between digital image and existing building model. Optimal building models for co-registration purpose are extracted as primitives from existing building model database. A set of homologous features between straight lines extracted from aerial digital image and model primitive are computed based on geometric similarity function. With obtained homologous features, EO parameter is recomputed using least square method. The result shows that die suggested method automatically co-register two data set in a reliable manner.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.42 no.1
• /
• pp.79-87
• /
• 2005

This paper presents an active contour model for extracting accurate facial regions in complex environments. In the model, a contour is represented by a zero level set of level function φ, and evolved via level set partial differential equations. Then, unlike general active contours, skin color information that is represented by 2D Gaussian model is used for evolving and slopping a curve, which allows the proposed method to be robust to noise and varying pose. To assess the effectiveness of the proposed method it was tested with several natural scenes, and the results were compared with those of geodesic active contours. Experimental results demonstrate the superior performance of the proposed method.

• The KIPS Transactions:PartB
• /
• v.10B no.5
• /
• pp.521-526
• /
• 2003

In this paper, we propose a geometric active contour model based on intensity information and curve evolution for detecting region boundaries. We put boundary extraction problem as the minimization of the difference between the average intensity of the region and the intensity of the expanding closed curves. We used level set theory to implement the curve evolution for optimal solution. It offered much more freedom in the initial curve position than a general active contour model. Our methods could detect regions whose boundaries are not necessarily defiened by gradient compared to general edge based methods and detect multiple boundaries at the same time. We could improve the result by using anisotropic diffusion filter in image preprocessing. The performance of our model was demonstrated on several data sets like CT and MRI medical images.

• Journal of the Korea Computer Industry Society
• /
• v.6 no.3
• /
• pp.437-446
• /
• 2005

Workflow mining is a newly emerging research issue for rediscovering and reengineering workflow models from workflow logs containing information about workflow being executed on the workflow engine. This paper newly defines a workflow process reduction mechanism that formally and automatically reduces an original workflow process to a minimal set of activities, which was used proposed 'E-walk series analysis algorithm'. Main purpose of this paper is to minimize discrepancies between the workflow process modeled and the enacted workflow process as it is actually being executed. That means, we compare a complete set of activity firing sequences on buildtime with workflow execution logs which was generate on runtime. For this purpose we proposed two algorithm, the one is 'Activity-Dependent Net Algorithm' and the other is 'E-Walk Series Analysis Algorithm'.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.3B
• /
• pp.421-430
• /
• 2010

Now a days, as a communications network is being broadband, IPTV(Internet Protocol Television) service which provides various two-way TV service is increasing. But as the data which is transmitted between IPTV set-top box and smart card is almost transmitted to set-top box, the illegal user who gets legal authority by approaching to the context of contents illegally using McComac Hack Attack is not prevented perfectly. In this paper, set-top box access security model is proposed which is for the protection from McComac Hack Attack that tries to get permission for access of IPTV service illegally making data line which is connected from smart card to set-top box by using same kind of other set-top box which illegal user uses. The proposed model reports the result of test which tests the user who wants to get permission illegally by registration the information of a condition of smart card which is usable in set-top box in certification server so that it prevents illegal user. Specially, the proposed model strengthen the security about set-top box by adapting public key which is used for establishing neighbor link and inter-certification process though secret value and random number which is created by Pseudo random function.

• International Journal of Highway Engineering
• /
• v.14 no.2
• /
• pp.11-17
• /
• 2012

This research describes how to predict a model of the tensile strain related to the fatigue cracking performance of several asphalt concrete structures through design of experiments(e.g., Response Surface Methodology) and harmony search(HS) algorithm. The axisymmetric analysis program of finite element method, which is the KICTPAVE, was used to determine the strain level at the interface layer between asphalt layer and lean concrete layer. Once the training database set of various strain levels was constructed under the several condition of layer stiffnesses and thicknesses in the asphalt concrete structures, the data set was trained through the HS algorithm in order to determine the regression coefficients defined based on a response surface methodology. Furthermore, the testing set, which was not used for the training procedure of HS algorithm, was also constructed in order to evaluate whether the regression coefficients of a prediction model can be appropriately applied for other cases in asphalt concrete structures.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.10 no.1
• /
• pp.85-94
• /
• 2017

In this paper, I proposed a sound model generation and a most frequent model search algorithm for recognizing animal vocalization. The sound model generation algorithm generates a optimal set of models through repeating processes such as the training process, the Viterbi Search process, and the most frequent model search process while adjusting HMM(Hidden Markov Model) structure to improve global recognition rate. The most frequent model search algorithm searches the list of models produced by Viterbi Search Algorithm for the most frequent model and makes it be the final decision of recognition process. It is implemented using MFCC(Mel Frequency Cepstral Coefficient) for the sound feature, HMM for the model, and C# programming language. To evaluate the algorithm, a set of animal sounds for 27 species were prepared and the experiment showed that the sound model generation algorithm generates 27 HMM models with 97.29 percent of recognition rate.