• 한국컴퓨터정보학회논문지
• /
• 제28권5호
• /
• pp.17-28
• /
• 2023

포장도로에서 발생하는 포트홀은 고속 주행 차량에 치명적인 영향을 미치며, 사망사고를 유발할 수 있는 도로상의 장애물이다. 이를 방지하기 위해 일반적으로는 작업자가 직접 포트홀을 탐지하는 방식을 사용해왔으나, 이는 작업자의 안전 문제와 예측하기 어려운 범주에서 발생하는 모든 포트홀을 인력으로 탐지하는 것이 비효율적이기 때문에 한계가 있다. 또한, 도로 환경과 관련된 지반 환경이 포트홀 생성에 영향을 미치기 때문에, 완벽한 포트홀 방지는 어렵다. 데이터셋 구축을 위해서는 전문가의 지도하에 라벨링 작업이 필요하지만, 이는 매우 시간과 비용이 많이 필요하다. 따라서, 본 논문에서는 Mean Teacher 기법을 사용하여 라벨링된 데이터의 샘플 수가 적더라도 지도학습보다 더욱 강인한 포트홀 이미지 분류 성능을 보여준다. 이러한 결과는 성능지표와 GradCAM을 통해 입증되었으며, 준지도학습을 사용할 때 15개의 사전 학습된 CNN 모델이 평균 90.41%의 정확도를 달성하며, 지도학습과 비교하여 2%에서 9%의 차이로 강인한 성능을 나타내는 것을 확인하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권11호
• /
• pp.4028-4042
• /
• 2021

Aiming at the problem of software defect prediction difficulty caused by insufficient software defect marker samples and unbalanced classification, a semi-supervised software defect prediction model based on a tri-training algorithm was proposed by combining feature normalization, over-sampling technology, and a Tri-training algorithm. First, the feature normalization method is used to smooth the feature data to eliminate the influence of too large or too small feature values on the model's classification performance. Secondly, the oversampling method is used to expand and sample the data, which solves the unbalanced classification of labelled samples. Finally, the Tri-training algorithm performs machine learning on the training samples and establishes a defect prediction model. The novelty of this model is that it can effectively combine feature normalization, oversampling techniques, and the Tri-training algorithm to solve both the under-labelled sample and class imbalance problems. Simulation experiments using the NASA software defect prediction dataset show that the proposed method outperforms four existing supervised and semi-supervised learning in terms of Precision, Recall, and F-Measure values.

• 한국지능시스템학회논문지
• /
• 제11권7호
• /
• pp.675-679
• /
• 2001

In this paper, we present a new competitive learning algorithm called Dynamic Competitive Learning (DCL). DCL is a supervised learning method that dynamically generates output neurons and initializes automatically the weight vectors from training patterns. It introduces a new parameter called LOG (Limit of Grade) to decide whether an output neuron is created or not. If the class of at least one among the LOG number of nearest output neurons is the same as the class of the present training pattern, then DCL adjusts the weight vector associated with the output neuron to learn the pattern. If the classes of all the nearest output neurons are different from the class of the training pattern, a new output neuron is created and the given training pattern is used to initialize the weight vector of the created neuron. The proposed method is significantly different from the previous competitive learning algorithms in the point that the selected neuron for learning is not limited only to the winner and the output neurons are dynamically generated during the learning process. In addition, the proposed algorithm has a small number of parameters, which are easy to be determined and applied to real-world problems. Experimental results for pattern recognition of remote sensing data and handwritten numeral data indicate the superiority of DCL in comparison to the conventional competitive learning methods.

• Management Science and Financial Engineering
• /
• 제16권1호
• /
• pp.81-117
• /
• 2010

Semi-supervised learning incorporates unlabeled examples, whose labels are unknown, as well as labeled examples into learning process. Although transductive support vector machine (TSVM), one of semi-supervised learning models, was proposed about a decade ago, its application to large-scaled data has still been limited due to its high computational complexity. Our previous research addressed this limitation by introducing a branch-and-bound algorithm for finding an optimal solution to TSVM. In this paper, we propose three new techniques to enhance the performance of the branch-and-bound algorithm. The first one tightens min-cut bound, one of two bounding strategies. Another technique exploits a graph-based approximation to a support vector machine problem to avoid the most time-consuming step. The last one tries to fix the labels of unlabeled examples whose labels can be obviously predicted based on labeled examples. Experimental results are presented which demonstrate that the proposed techniques can reduce drastically the number of subproblems and eventually computational time.

• 음성과학
• /
• 제10권1호
• /
• pp.71-84
• /
• 2003

In the acoustic modeling for large vocabulary speech recognition, a sparse data problem caused by a huge number of context-dependent (CD) models usually leads the estimated models to being unreliable. In this paper, we develop a new clustering method based on the C45 decision-tree learning algorithm that effectively encapsulates the CD modeling. The proposed scheme essentially constructs a supervised decision rule and applies over the pre-clustered triphones using the C45 algorithm, which is known to effectively search through the attributes of the training instances and extract the attribute that best separates the given examples. In particular, the data driven method is used as a clustering algorithm while its result is used as the learning target of the C45 algorithm. This scheme has been shown to be effective particularly over the database of low unknown-context ratio in terms of recognition performance. For speaker-independent, task-independent continuous speech recognition task, the proposed method reduced the percent accuracy WER by 3.93% compared to the existing rule-based methods.

• International Journal of Computer Science & Network Security
• /
• 제22권1호
• /
• pp.283-287
• /
• 2022

Machine Learning is the most popular method used in data science. Growth of data is not only numeric data but also text data. Most of the algorithm of supervised and unsupervised machine learning algorithms use numeric data. Now it is required to convert text data into numeric. There are many techniques for this conversion. Researcher confuses which technique is best in what situation. Here in proposed work BOW (Bag-of-Words) and TF-IDF (Term-Frequency-Inverse-Document-Frequency) has been studied based on different features to determine best method. After experimental results on text data, TF-IDF and BOW both provide better performance at range from 100 to 150 number of features.

• 한국통신학회논문지
• /
• 제30권2C호
• /
• pp.31-40
• /
• 2005

오류 역전파 알고리즘의 문제점과 ART 신경회로망의 문제점을 개선하기 위해 Jacobs가 제안한 delta-bar-delta 방법과 신경회로망을 결합한 자가 생성 지도 학습 알고리듬을 제안한다. 입력층과 은닉층에서는 ART-1과 ART-2 알고리듬을 이용하고, winner-take-all 방식은 완전 연결 구조이나 연결된 가중치만을 조정하도록 채택하였다. 실험을 위해 학생증, 주민등록증, 컨테이너의 영상으로 추출한 패턴을 신경회로망의 은닉층 노드에 대해 실험하였고, 실험결과 제안된 자기 생성 지도 학습알고리듬이 지역최소화, 학습 속도, 정체 현상이 기존의 방법보다 성능이 개선된 것을 확인하였다.

• 산업공학
• /
• 제24권1호
• /
• pp.8-14
• /
• 2011

Principal Component Analysis (PCA) reduces the dimensionality of the process by creating a new set of variables, Principal components (PCs), which attempt to reflect the true underlying process dimension. However, for highly nonlinear processes, this form of monitoring may not be efficient since the process dimensionality can't be represented by a small number of PCs. Examples include the process of semiconductors, pharmaceuticals and chemicals. Nonlinear correlated process variables can be reduced to a set of nonlinear principal components, through the application of Kernel Principal Component Analysis (KPCA). Support Vector Data Description (SVDD) which has roots in a supervised learning theory is a training algorithm based on structural risk minimization. Its control limit does not depend on the distribution, but adapts to the real data. So, in this paper proposes a non-linear process monitoring technique based on supervised learning methods and KPCA. Through simulated examples, it has been shown that the proposed monitoring chart is more effective than $T^2$ chart for nonlinear processes.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권10호
• /
• pp.3685-3707
• /
• 2021

In recent years, facial attribute editing has been successfully used to effectively change face images of various attributes based on generative adversarial networks and encoder-decoder models. However, existing models have a limitation in that they may change an unintended part in the process of changing an attribute or may generate an unnatural result. In this paper, we propose a model that improves the learning of the attention mask by adding a spatial attention mechanism based on the unified selective transfer network (referred to as STGAN) using semi-supervised learning. The proposed model can edit multiple attributes while preserving details independent of the attributes being edited. This study makes two main contributions to the literature. First, we propose an encoder-decoder model structure that learns and edits multiple facial attributes and suppresses distortion using an attention mask. Second, we define guide masks and propose a method and an objective function that use the guide masks for multiple facial attribute editing through semi-supervised learning. Through qualitative and quantitative evaluations of the experimental results, the proposed method was proven to yield improved results that preserve the image details by suppressing unintended changes than existing methods.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권7호
• /
• pp.2399-2413
• /
• 2021

A recommendation system is an information filter tool, which uses the ratings and reviews of users to generate a personalized recommendation service for users. However, the cold-start problem of users and items is still a major research hotspot on service recommendations. To address this challenge, this paper proposes a high-efficient hybrid recommendation system based on Fuzzy C-Means (FCM) clustering and supervised learning models. The proposed recommendation method includes two aspects: on the one hand, FCM clustering technique has been applied to the item-based collaborative filtering framework to solve the cold start problem; on the other hand, the content information is integrated into the collaborative filtering. The algorithm constructs the user and item membership degree feature vector, and adopts the data representation form of the scoring matrix to the supervised learning algorithm, as well as by combining the subjective membership degree feature vector and the objective membership degree feature vector in a linear combination, the prediction accuracy is significantly improved on the public datasets with different sparsity. The efficiency of the proposed system is illustrated by conducting several experiments on MovieLens dataset.