• Journal of Korean Society for Geospatial Information Science
• /
• v.7 no.1 s.13
• /
• pp.29-40
• /
• 1999

In vector GIS, natural linear entities (called linear entitles) are usually represented by a set of line segments. As an alternative of the line segments, curve segments can be used to represent the linear entities. The curve segments, as one-dimensional spatial objects, we generated by spline interpolation technique such as Bezier technique. In an effort to improve its accuracy in resembling the linear entities, the Bezier technique was modified generating a new technique (called New technique) (Kiyun, 1998). In this paper, validity of the New technique was tested. Test focused on answering two questions: (1) whether or not the curve segments from the New technique replace line segments so as to enhance the accuracy of representations of linear entities, and (2) whether or not the curve segments from the New technique represent the linear entities more accurately than curve segments from the Bezier technique. Answering these two questions entailed two hypothesis tests. For test data, a series of hydrologic lines on 7.5-minute USGS map series were selected. Test were done using t-test method and statistical inferences were made from the results. Test results indicated that curve segments from both the Bezier and New techniques represent the linear entities more accurately than the line segments do. In addition, curve segments from the New technique represent the linear entities more accurately than the line segments from the Bezier technique do at probability level 69% or higher.

• Journal of KIISE:Software and Applications
• /
• v.29 no.4
• /
• pp.270-277
• /
• 2002

Text filtering is a task of deciding whether a document has relevance to a specified topic. As Internet and Web becomes wide-spread and the number of documents delivered by e-mail explosively grows the importance of text filtering increases as well. The aim of this paper is to improve the accuracy of text filtering systems by using machine learning techniques. We apply AdaBoost algorithms to the filtering task. An AdaBoost algorithm generates and combines a series of simple hypotheses. Each of the hypotheses decides the relevance of a document to a topic on the basis of whether or not the document includes a certain word. We begin with an existing AdaBoost algorithm which uses weak hypotheses with their output of 1 or -1. Then we extend the algorithm to use weak hypotheses with real-valued outputs which was proposed recently to improve error reduction rates and final filtering performance. Next, we attempt to achieve further improvement in the AdaBoost's performance by first setting weights randomly according to the continuous Poisson distribution, executing AdaBoost, repeating these steps several times, and then combining all the hypotheses learned. This has the effect of mitigating the ovefitting problem which may occur when learning from a small number of data. Experiments have been performed on the real document collections used in TREC-8, a well-established text retrieval contest. This dataset includes Financial Times articles from 1992 to 1994. The experimental results show that AdaBoost with real-valued hypotheses outperforms AdaBoost with binary-valued hypotheses, and that AdaBoost iterated with random weights further improves filtering accuracy. Comparison results of all the participants of the TREC-8 filtering task are also provided.

• Journal of The Korean Association For Science Education
• /
• v.26 no.4
• /
• pp.568-580
• /
• 2006

The purpose of this study was to investigate the brain activities by 4-types of Generating Process of Scientific Emotion (GPSE) in the hypothesis-generating biological phenomena by using fMRI. Four-types of GPSE were involved in the Basic Generating Process (BGP), Retrospective Generating Process (RGP), Cognitive Generating Process (CGP) and Attributive Generating Process (AGP). For this study, we made an experimental design capable of validating the 4-types of generating process (e.g. BGP, RGP, CGP and AGP), and then measured BOLD signals of 10 pre-service teachers' brain activities by 3.0T fMRI system. Subjects were 10 healthy females majoring in biology education. As a result, there were clear differences among 4-types of GPSE. Brain areas activated by BGP were at right occipital lobe (BA 17), at left thalamus and left parahippocampal gyrus, while in the case of RGP, at left superior parietal lobe (BA 8, 9), at left pulvinar and left globus pallidus were activated. Brain areas activated by CGP were the right posterior cingulate and left medial frontal gyrus (BA 6). In the case of AGP, the most distinctively activated brain areas were the right medial frontal gyrus (BA 8) and left inferior parietal lobule (BA 40). These results would mean that each of the 4-types of GPSE has a specific neural networks in the brain, respectively. Furthermore, it would provide the basis of brain-based learning in science education.

• Journal of The Korean Association For Science Education
• /
• v.25 no.4
• /
• pp.503-513
• /
• 2005

The purpose of this study was to analyze the generating processes of scientific emotion, that appears during the generation of biological hypotheses. To perform the study, a tentative model was set up through pilot test, a think-aloud training procedure was planned and a standardized interview instrument was developed before getting protocols. In this study, 8 college students were selected to bring out protocol through the method of think-aloud, retrospective debriefing, focused interview and observing. As the result of analysis of the collected protocol through coding scheme, 4 types of process for scientific emotion-generating were sorted out. First type was a basic process which was a feeling process in prior to recognition. Second type was a retrospective process that explains the process of retrospect for emotional memory based on the past. Third type was a cognitive process and it explains emotion that occurs during thinking process to achieve cognitive goal. Fourth type was an attribution process and it explains that emotion is generated in the process of attribution for cognitive goal's achievement. These types of process of scientific emotion-generating can contribute the basis for developing cognitive model of EBL (Emotional Brain-based Learning) strategy.

• Journal of KIISE:Software and Applications
• /
• v.36 no.10
• /
• pp.861-873
• /
• 2009

There are many reasons that fail to get appropriate information in information retrieval. Allomorph is one of the reasons for search failure due to keyword mismatch. This research proposes a method to construct alternative word candidates automatically in order to minimize search failure due to keyword mismatch. Assuming that two words have similar meaning if they have similar co-occurrence words, the proposed method uses the concept of concentration, association word set, cosine similarity between association word sets and a filtering technique using confidence. Performance of the proposed method is evaluated using a manually extracted alternative list. Evaluation results show that the proposed method outperforms the context window overlapping in precision and recall.

• Korean Journal of Cognitive Science
• /
• v.19 no.1
• /
• pp.57-74
• /
• 2008

This study aimed to investigate the preferential brain activations involed in the set size during random number generation (RNG). The BNG condition gave more increased activations in the anterior cingulate cortex (ACC), inferior frontal gyrus (IFG), inferior parietal lobule (IPL), and superior temporal gyrus (STG) than the simple counting condition, which was a control rendition. When the activations were compared by the small set size condition versus the large set size condition, broad areas covering tempore-occipital network, ACC, and postcentral gyrus were more highly activated in the small set size condition than in the large set size condition, while responses of areas including medial frontal gyrus, superior parietal lobule, and lingual gyrus were more increased in the large set size condition than in the small set size condition. The capacity hypothesis of working memory fails to explain the results. On the contrary, strategy selection hypothesis seems to explain the current observations properly.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.8A
• /
• pp.639-647
• /
• 2012

Vision-based on-road vehicle detection is one of the key techniques in automotive driver assistance systems. However, due to the huge within-class variability in vehicle appearance and environmental changes, it remains a challenging task to develop an accurate and reliable detection system. In general, a vehicle detection system consists of two steps. The candidate locations of vehicles are found in the Hypothesis Generation (HG) step, and the detected locations in the HG step are verified in the Hypothesis Verification (HV) step. Since the final decision is made in the HV step, the HV step is crucial for accurate detection. In this paper, we propose using a reduced multivariate polynomial pattern classifier (RM) for the HV step. Our experimental results show that the RM classifier outperforms the well-known Support Vector Machine (SVM) classifier, particularly in terms of the fast decision speed, which is suitable for real-time implementation.

• Journal of the Institute of Convergence Signal Processing
• /
• v.11 no.4
• /
• pp.345-353
• /
• 2010

A general solution in global self-position location of robot is to generate multiple hypothesis in self-position of robot, which is to look for the most positive self-position by evaluating each hypothesis based on features of observed landmark. Markov Localization(ML) or Monte Carlo Localization(MCL) to be the existing typical method is to evaluate all pairs of landmark features and generated hypotheses, it can be said to be an optimal method in sufficiently calculating resources. But calculating quantities was proportional to the number of pairs to evaluate in general, so calculating quantities was piled up in wide environments in the presence of multiple pairs if using these methods. First of all, the positive and promising pairs is located and evaluated to solve this problem in this paper, and the newly locating method to make effective use of calculating time is proposed. As the basic method, it is used both RANSAC(RANdom SAmple Consensus) algorithm and preemption scheme to be efficiency method of RANSAC algorithm. The calculating quantity on each observation of robot can be suppressed below a certain values in the proposed method, and the high location performance can be determined by an experimental on verification.

• The Korean Journal of Financial Management
• /
• v.18 no.1
• /
• pp.1-21
• /
• 2001

주가가 장기기억과정에 의하여 생성되면 주가과정에 가해진 충격은 쌍곡선감소율로 소멸한다. 따라서 충격의 영향이 대단히 느리게 감소하여 충격이 지속성을 가진다. 반면 주가가 단기 기억과정을 따르면 지수율로 감소하여 소멸한다. 지수율감소는 충격의 영향을 급속히 소멸시키므로 충격의 영향이 조만간 소멸한다. 따라서 충격으로 변화된 주가는 평균으로 회귀한다. 충격의 영향이 영원히 존재하는 과정도 존재한다. 장기기억과정은 쪽거리차분과정 또는 분수차분과정이다. 차분모수가 분수일 것이 요구되는 시계열은 장기기억과정이다. 주가가 장기기억과정에 의하여 생성되고 있는지의 여부를 검정하였다. 장기기억과정을 형성시키는 차분모수는 분수차분모수이다. 일별 주가지수의 수익률을 사용하여 차분모수를 추정하였는 바 그 값이 0에 근접하고 있음이 밝혀졌다. 그러나 Kospi, Nasdaq과 Mib30은 장기기억모수가 0에 접근하고 있으나 0이 아니다. 따라서 이 지수들은 장기기억과정에 의하여 생성된다고 할 수 있다. 반면 Dow Jones, S&P 500와 Dax는 장기기억모수가 0이라는 가설이 기각되지 않고 있어 이 지수들은 단기기억과정을 따르고 있다. 따라서 평균회귀과정에 의하여 생성되고 있음을 알 수 있다.

• Korean Journal of Cognitive Science
• /
• v.3 no.1
• /
• pp.79-111
• /
• 1991

This paper suggests a system which analyzes facial images through hypotheses generation and verification.A model knowledge on faical structures is to be represented in a hieratchical frame system and the process of hypothesis generation and verification is to be embodied through linking upper lower-level frames.This paper especially addresses the problem of how to selec an approproate knowledge at each stage of hypotheses generation and verification.