• Proceedings of the Korean Information Science Society Conference
• /
• 2002.04b
• /
• pp.154-156
• /
• 2002

정보화시대에 들어서면서 나날이 급증하는 데이터에 대한 재가용성을 위한 많은 연구가 이루어지고 있다 이러한 연구들은 의사결정지원, 예측, 추정 등의 분야에서 적용되고 있으나, 실생활에 활발히 적용되기까지 앞으로 많은 연구 및 개발이 요구된다. 이 논문에서는 수집된 데이터로부터 패턴을 추출하여 예측결과를 제공할 수 있는 시스템 모델과 모델에 적합한 점진적 규칙갱신 알고리즘을 제안하였다. 제안하는 예측 모델의 특징은 새로 입력되는 정보에 대한 반복 학습시 수치데이터에 대한 평균근사치 할당방법을 적용하여 규칙갱신을 용이하게 하였으며 각 클래스의 수치데이터에 대한 분류를 용이하도록 하였다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10a
• /
• pp.190-192
• /
• 2003

소비자의 신용 대출 규모가 점차 증가하면서 기업에서 고객의 신용 등급에 의한 정확한 고객 분류를 필요로 하고 있다 이를 위해 판별 분석과 신경망의 역전파(BP: Back Propagation)를 이용한 고객 분류 시스템이 연구되었다. 그러나, 판별 분석을 사용한 방법은 불규칙한 신용 거래의 성향을 보이는 비정규 분포의 고객 데이터의 영향으로 여러 개의 판별 함수와 판별점이 존재하여 분류 정확도가 떨어지는 단점이 있다. 신경망을 이용한 방법은 불규칙한 신용 거래의 성향을 보이는 고객 데이터에 의해서, 지역 최소점(Local Minima)에 빠져 최대의 분류 정확률을 보이는 분류자를 얻지 못하는 경우가 발생할 수 있다. 본 논문에서는 이러한 기존 연구의 분류 정확률을 저하시키는 단점을 해결하기 위해 SVM(Support Vector Machine)을 사용하여 고객의 신용 등급을 분류하는 방법을 제안한다. SVM은 SV(Support Vector)의 수에 의해서 학습 성능이 좌우되므로, 불규칙한 거래 성향을 보이는 고객에 대해서도 높은 차원으로의 매핑을 통하여, 효과적으로 학습시킬 수 있어 분류의 정확도를 높일 수 있다 하지만, SVM은 근사화 알고리즘(Approximation Algorithms)을 이용하므로 분류 정확도가 이론적인 성능에 미치지 못한다. 따라서, 본 논문은 점진적 앙상블 SVM을 사용하여, 기존의 고객 분류 시스템의 문제점을 해결하고 실제적으로 SVM의 분류 정확률을 높인다. 실험 결과는 점진적 앙상블 SVM을 이용한 방법의 정확성이 기존의 방법보다 높다는 것을 보여준다.

• Journal of KIISE:Software and Applications
• /
• v.34 no.2
• /
• pp.112-122
• /
• 2007

Training a statistical model for semantic role labeling requires a large amount of manually tagged corpus. However. such corpus does not exist for Korean and constructing one from scratch is a very long and tedious job. This paper suggests a modified algorithm of self-training, an unsupervised algorithm, which trains a semantic role labeling model from any raw corpora. For initial training, a small tagged corpus is automatically constructed iron case frames in Sejong Electronic Dictionary. Using the corpus, a probabilistic model is trained incrementally, which achieves 83.00% of accuracy in 4 selected adverbial cases.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.9
• /
• pp.1979-1986
• /
• 2010

Biological early warning system detects toxicity by looking at behavior of organisms in water. The system uses classifier for judgement about existence and amount of toxicity in water. Boosting algorithm is one of possible application method for improving performance in a classifier. Boosting repetitively change training example set by focusing on difficult examples in basic classifier. As a result, prediction performance is improved for the events which are difficult to classify, but the information contained in the events which can be easily classified are discarded. In this paper, an incremental learning method to overcome this shortcoming is proposed by using the extended data expression. In this algorithm, decision tree classifier define class distribution information using the weight parameter in the extended data expression by exploiting the necessary information not only from the well classified, but also from the weakly classified events. Experimental results show that the new algorithm outperforms the former Learn++ method without using the weight parameter.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.18 no.1
• /
• pp.113-117
• /
• 2018

Human emotions are revealed by various factors. Words, actions, facial expressions, attire and so on. But people know how to hide their feelings. So we can not easily guess its sensitivity using one factor. We decided to pay attention to behaviors and facial expressions in order to solve these problems. Behavior and facial expression can not be easily concealed without constant effort and training. In this paper, we propose an algorithm to estimate human emotion through combination of two results by gradually learning human behavior and facial expression with little data through the deep learning method. Through this algorithm, we can more comprehensively grasp human emotions.

• Annual Conference on Human and Language Technology
• /
• 2006.10e
• /
• pp.32-39
• /
• 2006

본 논문은 한국어정보처리 과정에서 구문 관계를 의미 관계로 사상하는 의미역 결정 문제에 대해 다루고 있다. 한국어의 경우 대량의 학습 말뭉치를 구하기 힘들며, 이를 구축하기 위해서는 많은 시간과 노력이 필요한 문제점이 있다. 따라서 본 논문에서는 학습 말뭉치를 직접 태깅하지 않고 격틀사전을 이용하여 자동으로 학습 말뭉치를 구축하고 간단한 확률모델을 적용하여 점진적으로 모델을 학습하는 수정된 self-training 알고리즘을 사용하였다. 실험 결과, 4개의 부사격 조사에 대해 평균적으로 81.81%의 정확률을 보였으며, 수정된 self-training 방법은 기존의 방법에 비해 성능 및 실행시간에서 개선된 결과를 보였다.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.9
• /
• pp.77-81
• /
• 2016

Logistic regression (LR) has been widely used for predicting the relationships among variables in various fields. We propose a new logistic regression model with a fine-grained weighting method, called value weighted logistic regression, by assigning different weights to each feature value. A gradient approach is utilized to obtain the optimal weights of feature values. We conduct experiments on several data sets and the experimental results show that the proposed method shows meaningful improvement in prediction accuracy.

• Journal of KIISE:Software and Applications
• /
• v.31 no.3
• /
• pp.379-386
• /
• 2004

Video segmentation is a fundamental task in video indexing and it includes two kinds of shot change detections such as the abrupt transition and the gradual transition. The abrupt shot boundaries are detected by computing the image-based distance between adjacent frames and comparing this distance with a pre-determined threshold value. However, the gradual shot boundaries are difficult to detect with this approach. To overcome this difficulty, we propose the method that detects gradual transition in the MPEG compressed video using the HMM (Hidden Markov Model). We take two different HMMs such as a discrete HMM and a continuous HMM with a Gaussian mixture model. As image features for HMM's observations, we use two distinct features such as the difference of histogram of DC images between two adjacent frames and the difference of each individual macroblock's deviations at the corresponding macroblock's between two adjacent frames, where deviation means an arithmetic difference of each macroblock's DC value from the mean of DC values in the given frame. Furthermore, we obtain the DC sequences of P and B frame by the first order approximation for a fast and effective computation. Experiment results show that we obtain the best detection and classification performance of gradual transitions when a continuous HMM with one Gaussian model is taken and two image features are used together.

• School Mathematics
• /
• v.7 no.4
• /
• pp.403-425
• /
• 2005

The study aims to get real picture of primary mathematics education based on RME in the Netherlands focusing on number and operations strand by reflecting and analyzing the documents in relation to the primary school mathematics curriculum. In order to attain these purposes, the present paper describes the core goals for mathematics education, Dutch Pluspunt textbook series for the primary school, and a learning-teaching trajectory by TAL project which are determinants of the Dutch primary school mathematics curriculum. Under these reflections on the documents, it is analyzed what is the characteristics of number and operations strand in the Nether-lands as follows: counting numbers, contextualization, positioning, structuring, progressive algoritmization based on levels, estimation and insightful use of a calculator. Finally, discussing Points for improving our primary mathematics curriculum and textbook series development are described.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.9A
• /
• pp.1411-1418
• /
• 1999

In this paper, we propose a fast motion estimation algorithm using motion prediction and neural network. Considering that the motion vectors have high spatial correlation, the motion vector of current block is predicted by those of neighboring blocks. The codebook of motion vector is designed by Kohonen self-organizing feature map(KSFM) learning algorithm which has a fast learning speed and 2-D adaptive chararteristics. Since the similar codevectors are closely located in the 2-D codebook the motion is progressively estimated from the predicted codevector in the codebook. Computer simulation results show that the proposed method has a good performance with reduced computational complexity.