• 경영정보학연구
• /
• 제1권2호
• /
• pp.55-71
• /
• 1999

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권11호
• /
• pp.5427-5445
• /
• 2019

While deep neural networks have achieved remarkable performance in representation learning, a huge amount of labeled training data are usually required by supervised deep models such as convolutional neural networks. In this paper, we propose a new representation learning method, namely generative adversarial networks (GAN) based bagging deep convolutional autoencoders (GAN-BDCAE), which can map data to diverse hierarchical representations in an unsupervised fashion. To boost the size of training data, to train deep model and to aggregate diverse learning machines are the three principal avenues towards increasing the capabilities of representation learning of neural networks. We focus on combining those three techniques. To this aim, we adopt GAN for realistic unlabeled sample generation and bagging deep convolutional autoencoders (BDCAE) for robust feature learning. The proposed method improves the discriminative ability of learned feature embedding for solving subsequent pattern recognition problems. We evaluate our approach on three standard benchmarks and demonstrate the superiority of the proposed method compared to traditional unsupervised learning methods.

• 한국정보통신학회:학술대회논문집
• /
• 한국해양정보통신학회 1999년도 춘계종합학술대회
• /
• pp.179-184
• /
• 1999

디지털 시스템을 설계하고 분석할 때, 번지버스와 데이터버스 및 각종 제어 신호들을 타이밍도로 표현하는 경우가 많다. 그러나 디지털 시스템의 동작이 타이밍도로 표현되는 경우, 그 표현이 복잡할 뿐 아니라 동작 분석이 용이하지 못하다. 본 연구에서는 시스템의 타이밍도를 확장된 마크흐름선도(EMFG; Extended Mark flow Graph)로 표현하는 방법을 제안하였다. 시스템의 동작이 EMFG로 표현되는 경우, 각종 신호들에 따라 변화하는 시스템의 상태가 도식적으로 표현되므로 시스템의 동작 분석이 용이해질 뿐 아니라 시스템의 설계에도 유용하게 이용 될 수 있다. 적용 예로 NEC사의 $\mu$PD70320 CPU의 메모리 읽기 사이클 및 MCM60256A의 메모리 동작을 EMFG로 표현하였다.

• 한국정보통신학회논문지
• /
• 제8권5호
• /
• pp.1089-1095
• /
• 2004

주어진 네트워크에서 최적의 스타이너 트리를 구하는 문제는 NP-hard이며, 최적에 가까운 스타이너 트리를 구하기 위하여 유전자 알고리즘을 이용한다. 본 논문에서는 이 문제를 해결하기 위하여 유전자 알고리즘에서 염색체를 기존의 이진스트링 대신 부동소수점으로 표현하였다. 먼저 주어진 네트워크에 Prim의 알고리즘을 적용하여 스패닝 트리를 구하고, 부동소수점 표현을 갖는 유전자 알고리즘을 사용하여 새로운 스타이너 점을 트리에 추가하는 과정을 반복함으로써 최적에 가까운 스타이너 트리를 구했다 이 방법을 사용하면 이진스트링을 사용하는 기존의 방법에 비해서 트리가 보다 빠르고 정확하게 최적에 가까운 스타이너 트리에 접근했다.

• 지능정보연구
• /
• 제17권2호
• /
• pp.1-22
• /
• 2011

본 논문에서는 선행제약순서결정문제(Sequencing problem with precedence constraints, SPPC)를 효과적으로 해결하기 위한 적응형 유전알고리즘(Adaptive genetic algorithm, aGA)을 제안한다. aGA에서 는 SPPC를 효과적으로 표현하기 위해 위상정렬에 기초한 표현절차(topological sort-based representation procedure) 를 사용한다. 제안된 aGA는 퍼지로직제어를 이용한 적응형구조를 가지고 있으며, 유전 탐색과정을 통해 교차변이 연산자(Crossover operator)의 비율을 적응적으로 조절한다. 수치예제에서는 다양한 형태의 SPPC를 제시하였으며, 그 실험결과는 제안된 aGA가 기존의 알고리즘보다 우수함을 보여주었다. 결론적으로 말하자면 본 논문에서는 제안된 aGA가 다양한 형태의 SPPC에서 최적해 혹은 최적순서를 발견하는데 아주 효과적이라는 것을 밝혔다.

• Journal of Information Processing Systems
• /
• 제16권2호
• /
• pp.261-276
• /
• 2020

Digital contents services are one of the topics that have been intensively studied in the media industry, where various semantic and ontology techniques are applied. However, query execution for ontology data is still inefficient, lack of sufficient extensible definitions for node relationships, and there is no specific semantic method fit for media data representation. In order to make the machine understand digital contents (DCs) data well, we analyze DCs data, including static data and dynamic data, and use ontology to specify and classify objects and the events of the particular objects. Then the formal representation method is proposed which not only redefines DCs data based on the technology of OWL/RDF, but is also combined with media segmentation methods. At the same time, to speed up the access mechanism of DCs data stored under the persistent database, an ontology-based DCs query solution is proposed, which uses the specified distance vector associated to a surveillance of semantic label (annotation) to detect and track a moving or static object.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제10권7호
• /
• pp.3272-3285
• /
• 2016

Recently, Vector of Locally Aggregated Descriptors (VLAD) has been proposed to index image by compact representations, which encodes powerful local descriptors and makes significant improvement on search performance with less memory compared against the state of art. However, its performance relies heavily on the size of the codebook which is used to generate VLAD representation. It indicates better accuracy needs higher dimensional representation. Thus, more memory overhead is needed. In this paper, we enhance VLAD image representation by using two level hierarchical-codebooks. It can provide more accurate search performance while keeping the VLAD size unchanged. In addition, hierarchical-codebooks are used to construct multiple inverted files for more accurate non-exhaustive search. Experimental results show that our method can make significant improvement on both VLAD image representation and non-exhaustive search.

• Journal of the Optical Society of Korea
• /
• 제20권6호
• /
• pp.752-761
• /
• 2016

Hyperspectral images are often contaminated with stripe noise, which severely degrades the imaging quality and the precision of the subsequent processing. In this paper, a variational model is proposed by employing spectral-spatial adaptive unidirectional variation and a sparse representation. Unlike traditional methods, we exploit the spectral correction and remove stripes in different bands and different regions adaptively, instead of selecting parameters band by band. The regularization strength adapts to the spectrally varying stripe intensities and the spatially varying texture information. Spectral correlation is exploited via dictionary learning in the sparse representation framework to prevent spectral distortion. Moreover, the minimization problem, which contains two unsmooth and inseparable $l_1$-norm terms, is optimized by the split Bregman approach. Experimental results, on datasets from several imaging systems, demonstrate that the proposed method can remove stripe noise effectively and adaptively, as well as preserve original detail information.

• Journal of Information Technology Applications and Management
• /
• 제25권2호
• /
• pp.1-21
• /
• 2018

This study aims to investigate the usage patterns of users in Social Network Services (SNS) where is an upsurge. Specifically, the paper considers the reason why young people more and more prefer online (or mobile) SNS activities rather than offline face-to-face social relationship. Furthermore, the drivers which affect SNS usages are considered from users' personal characteristics and social influences. User's personal characteristics include their personalities (extraversion and introversion), narcissism, and life satisfaction. Social influences involve subjective norm, visibility, and image. Affected by personal and social factors in SNS, users intend to show positive self-presentation, which refers to a behavior to selectively expose his/her goodness to others. As one of the most influential drivers affecting SNS usage, the positive self-representation has an effect on the level of SNS usage. Thus, this paper suggests the hypothesized research model focusing on positive self-representation in the relationship among personal characteristics, social influence, and user's behavior in SNS. Empirical data analysis with 100 questionnaires suggests that all hypotheses were adopted except for the effect of visibility among social influence factors on positive self-representation.

• 한국경영과학회:학술대회논문집
• /
• 대한산업공학회/한국경영과학회 1996년도 춘계공동학술대회논문집; 공군사관학교, 청주; 26-27 Apr. 1996
• /
• pp.421-424
• /
• 1996

To provide an appropriate knowledge representation technique dependent on a particular domain, we consider the combine analytic hierachy process(CAHP). This is an extended method of the conventional AHP which is useful when two different expert groups are involved. Our problem domain is confined to human resource management including such major activities as planning, selection, placement, compensations, performance evaluation, training, and labor-management relations. We prioritize rules, frames, semantic nets, and predicate logic representation techniques best suited to each and all domains through an exploratory study.