• Journal of the Korean Society of Propulsion Engineers
• /
• v.10 no.2
• /
• pp.102-109
• /
• 2006

In this Paper, Support Vector Machine(SVM) and Artificial Neural Network(ANN) are used for developing the defect diagnostic algorithm of the aircraft turbo-shaft engine. The system that uses the ANN falls in a local minima when it learns many nonlinear data, and its classification accuracy ratio becomes low. To make up for this risk, the Separate Learning Algorithm(SLA) of ANN has been proposed by using SVM. This is the method that ANN learns selectively after discriminating the defect position by SVM, then more improved performance estimation can be obtained than using ANN only. The proposed SLA can make the higher classification accuracy by decreasing the nonlinearity of the massive data during the training procedure.

• Journal of KIISE:Software and Applications
• /
• v.34 no.1
• /
• pp.11-17
• /
• 2007

One of the popular methods used for pattern classification is the MBR (Memory-Based Reasoning) algorithm. Since it simply computes distances between a test pattern and training patterns or hyperplanes stored in memory, and then assigns the class of the nearest training pattern, it cannot explain how the classification result is obtained. In order to overcome this problem, we propose an incremental teaming algorithm based on RPA (Recursive Partition Averaging) to extract IF-THEN rules that describe regularities inherent in training patterns. But rules generated by RPA eventually show an overfitting phenomenon, because they depend too strongly on the details of given training patterns. Also RPA produces more number of rules than necessary, due to over-partitioning of the pattern space. Consequently, we present the IREA (Incremental Rule Extraction Algorithm) that overcomes overfitting problem by removing useless conditions from rules and reduces the number of rules at the same time. We verify the performance of proposed algorithm using benchmark data sets from UCI Machine Learning Repository.

• Smart Media Journal
• /
• v.9 no.4
• /
• pp.9-16
• /
• 2020

Recently, various studies are being conducted to integrate Image Segmentation into smart factory industries and autonomous driving fields. In particular, Image Segmentation systems using deep learning algorithms have been researched and developed enough to learn from large volumes of data with higher accuracy. In order to use image segmentation in the autonomous driving sector, sufficient amount of learning is needed with large amounts of data and the streaming environment that processes drivers' data in real time is important for the accuracy of safe operation through highways and child protection zones. Therefore, we proposed a novel DFCN algorithm that enhanced existing FCN algorithms that could be applied to various road environments, demonstrated that the performance of the DFCN algorithm improved 1.3% in terms of "loss" value compared to the previous FCN algorithms. Moreover, the proposed DFCN algorithm was applied to the existing U-Net algorithm to maintain the information of frequencies in the image to produce better results, resulting in a better performance than the classical FCN algorithm in the autonomous environment.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.93-93
• /
• 2023

최근 범지구적인 기후변화로 인해 도시유역의 홍수 발생 빈도가 빈번하게 발생하고 있다. 이로 인해 불투수성이 큰 도시지역의 침수 등의 자연재해 증가로 인명 및 재산피해가 발생하고 있다. 이에 따라 하수도의 제 기능을 수행하고 있다면 문제가 없지만 이상기후로 인한 기록적인 폭우에 의해 침수가 발생하고 있다. 홍수 및 집중호우와 같은 극치사상의 발생빈도가 증가됨에 따라 강우 사상의 변동에 따른 하수관로의 수위를 예측하고 침수에 대해 대처하기 위해 과거 수위에 따른 수위 예측은 중요할 것으로 판단된다. 본 연구에서는 수위 예측 연구에 많이 활용되고 있는 시계열 학습에 탁월한 LSTM 알고리즘을 활용한 하수관로 수위 예측을 진행하였다. 데이터의 학습과 검증을 수행하기 위해 실제 하수관로 수위 데이터를 수집하여 연구를 수행하였으며, 대상자료는 서울특별시 강동구에 위치한 하수관로 수위 자료를 활용하였다. 하수관로 수위 예측에는 딥러닝 알고리즘 RNN-LSTM 알고리즘을 활용하였으며, RNN-LSTM 알고리즘은 하천의 수위 예측에 우수한 성능을 보여준 바 있다. 1분 뒤 하수관로 수위 예측보다 5분, 10분 뒤 또는 1시간 3시간 등 다양한 분석을 실시하였다. 데이터 분석을 위해 하수관로 수위값 변동이 심한 1주일을 선정하여 분석을 실시하였다. 연구에는 Google에서 개발한 딥러닝 오픈소스 라이브러리인 텐서플로우를 활용하였으며, 하수관로 수위 고유번호 25-0001을 대상으로 예측을 하였다. 학습에는 2012년 ~ 2018년의 하수관로 수위 자료를 활용하였으며, 모형의 검증을 위해 결정계수(R square)를 이용하여 통계분석을 실시하였다.

• Journal of the Korea Computer Graphics Society
• /
• v.22 no.3
• /
• pp.21-29
• /
• 2016

Cell segmentation is an important but time-consuming and laborious task in biological image analysis. An automated, robust, and fast method is required to overcome such burdensome processes. These needs are, however, challenging due to various cell shapes, intensity, and incomplete boundaries. A precise cell segmentation will allow to making a pathological diagnosis of tissue samples. A vast body of literature exists on cell segmentation in microscopy images [1]. The majority of existing work is based on input images and predefined feature models only - for example, using a deformable model to extract edge boundaries in the image. Only a handful of recent methods employ data-driven approaches, such as supervised learning. In this paper, we propose a novel data-driven cell segmentation algorithm for bright-field microscopy images. The proposed method minimizes an energy formula defined by two dictionaries - one is for input images and the other is for their manual segmentation results - and a common sparse code, which aims to find the pixel-level classification by deploying the learned dictionaries on new images. In contrast to deformable models, we do not need to know a prior knowledge of objects. We also employed convolutional sparse coding and Alternating Direction of Multiplier Method (ADMM) for fast dictionary learning and energy minimization. Unlike an existing method [1], our method trains both dictionaries concurrently, and is implemented using the GPU device for faster performance.

• Journal of KIISE:Software and Applications
• /
• v.31 no.9
• /
• pp.1161-1170
• /
• 2004

This paper proposes to segment skin color areas using a clustering algorithm. Most of previously proposed clustering algorithms have some difficulties, since they generally detect hyperspherical clusters, run in a batch mode, and predefine a number of clusters. In this paper, we use a well-known elliptical clustering algorithm, an EM algorithm, and modify it to learn on-line and find automatically the number of clusters, called to an EAM algorithm. The effectiveness of the EAM algorithm is demonstrated on a task of skin color region segmentation. Experimental results present the EAM algorithm automatically finds a right number of clusters in a given image without any information on the number. Comparing with the EM algorithm, we achieved better segmentation results with the EAM algorithm. Successful results were achieved to detect and segment skin color regions using a conditional probability on a region. Also, we applied to classify images with persons and got good classification results.

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.11
• /
• pp.2897-2907
• /
• 1998

The production-rule generation from training examples is a hard problem that has large space and many local optimal solutions. Many learning methods are proposed for production-rule generation and genetic algorithms is an alternative learning method. However, traditional genetic algorithms has been known to have an obstacle in converging at the global solution area and show poor efficiency of production-rules generated. In this paper, we propose a production-rule generating method which uses genetic algorithm learning. By analyzing optimal sub-solutions captured by genetic algorithm learning, our method takes advantage of its schema structure and thus generates relatively small rule set.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.39 no.4
• /
• pp.462-469
• /
• 2002

While the conventional K-means algorithms use a fixed weight to design a vector codebook for all learning iterations, the proposed method employs a variable weight for learning iterations. The weight value of two or more beyond a convergent region is applied to obtain new codevectors at the initial learning iteration. The number of learning iteration applying a variable weight must be decreased for higher weight value at the initial learning iteration to design a better codebook. To enhance the splitting method that is used to generate an initial codebook, we propose a new method, which reduces the error between a representative vector and the member of training vectors. The method is that the representative vector with maximum squared error is rejected, but the vector with minimum error is splitting, and then we can obtain the better initial codevectors.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.8B
• /
• pp.1490-1499
• /
• 1999

This paper proposes a new handwritten numeral recognition method using numeral shape grasps and FSOM (Fuzzy Self-Organizing Map). The proposed algorithm is based on the idea that numeral input data with similar shapes are classified into the same class. Shapes of numeral data are created using lines of external-contact and the class of numeral data is determined by template matching of the shapes. Each class of numeral data has FSOM and feature extraction method, respectively. In this paper, we divide the numeral database into the 16 classes. The divided FSOM model allows not only an independent learning phase of SOM but also step-by-step learning. Experiments using Concordia University handwritten numeral database proved that the proposed algorithm is effective to improve recognition accuracy.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.05a
• /
• pp.141-144
• /
• 2008

본 논문에서는 세포영상을 분할하고 분류하는 알고리즘을 제안한다. 우선, 배경으로부터 세포를 분할한 후, 학습데이터로부터 얻은 Compactness, Smoothness, Moments와 같은 형태학적 특징을 추출한다. 전경세포들이 분할된 후에, 보다 정밀한 세포분석을 위해서 군집세포(Overlapped Cell)와 독립세포(Isolated Cell)를 분류 할 수 있는 알고리즘의 개발이 필수적이다. 이를 위해서 본 논문에서는 베이지안 네트워크와 각 노드에 대한 3개의 확률밀도함수를 사용하여 각 세포 영역을 분류한다. 분류된 군집세포영역은 향후 정확한 세포 분석을 위해서 군집세포가 포함하는 독립세포의 수만큼 마커를 찾고, Watershed 알고리즘과 병합과정을 거쳐 하나의 독립세포를 분리하게 된다. 현미경으로부터 얻은 세포영상에 대한 실험 결과는 이전 논문들에서 제안한 방법들과 비교했을 때, 각 군집세포의 독립세포로의 분리 이전에 세포영역에 대한 분류과정을 먼저 수행하였기 때문에 분할 성능이 크게 향상되었음을 확인할 수 있다.