• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.466-468
• /
• 2000

이 논문에서는 Helmholtz machine을 사용하여 데이터의 분포 추정을 함으로써 문서 분류기를 학습하는 방법 제안한다. Helmholtz machine 은 생성 모델과 인식 모델로 구성된 그래프 모델로서, 그래프 모델에서의 분포 추정을 보다 가능하게 하기 위한 근사 방법 중의 하나이다. Helmholtz machine에서의 각 입력 노드는 문서를 구성하는 하나의 단어에 대응하는 이진 노드이다. 입력 노드의 개수가 많아지면 그만큼 학습 시간이 증가하기 때문에, 학습 시간을 줄이면서 적정 수준의 성능을 유지하기 위해 자질 선정이 필요하다. 이러한 요구 사항을 충족시키기 위해 정보획득량(information gain)기준을 이용하였으며, 뉴스 그룹 데이터에 대해 그 성능을 측정하고 Naive Bayes를 이용한 것과 비교한다.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.13 no.2
• /
• pp.102-110
• /
• 2017

Current word sense disambiguation techniques employ various machine learning-based methods. Various approaches have been proposed to address this problem, including the knowledge base approach. This approach defines the sense of an ambiguous word in accordance with knowledge base information with no training corpus. In unsupervised learning techniques that use a knowledge base approach, graph-based and similarity-based methods have been the main research areas. The graph-based method has the advantage of constructing a semantic graph that delineates all paths between different senses that an ambiguous word may have. However, unnecessary semantic paths may be introduced, thereby increasing the risk of errors. To solve this problem and construct a fine-grained graph, in this paper, we propose a model that iteratively constructs the graph while eliminating unnecessary nodes and edges, i.e., senses and semantic paths. The hybrid similarity estimation model was applied to estimate a more accurate sense in the constructed semantic graph. Because the proposed model uses BabelNet, a multilingual lexical knowledge base, the model is not limited to a specific language.

• The Journal of the Korea Contents Association
• /
• v.16 no.2
• /
• pp.374-386
• /
• 2016

Provenance means the meta data that represents the history or lineage of a data in collaboration storage environments. Therefore, as provenance has been accruing over time, it takes several ten times as large as the original data. The schemes for effciently compressing huge amounts of provenance are required. In this paper, we propose a provenance compression scheme considering the RDF graph patterns. The proposed scheme represents provenance based on a standard PROV model and encodes provenance in numeric data through the text encoding. We compress provenance and RDF data using the graph patterns. Unlike conventional provenance compression techniques, we compress provenance by considering RDF documents on the semantic web. In order to show the superiority of the proposed scheme, we compare it with the existing scheme in terms of compression ratio and the processing time.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.21 no.7
• /
• pp.1327-1334
• /
• 2017

The Bubble sort graph is node symmetric, and can be used in the data sorting algorithm. In this research we propose and analyze that Half Bubble sort graph that improved the network cost of Bubble sort graph. The Half Bubble sort graph's number of node is n!, and its degree is ${\lfloor}n/2{\rfloor}+1$. The Half Bubble sort graph's degree is $${\sim_=}0.5$$ times of the Bubble sort, and diameter is $${\sim_=}0.9$$ times of the Bubble sort. The network cost of the Bubble sort graph is $${\sim_=}0.5n^3$$, and the network cost of the half Bubble sort graph is $${\sim_=}0.2n^3$$. We have proved that half bubble sort graph is a sub graph of the bubble sort graph. In addition, we proposed a routing algorithm and analyzed the diameter. Finally, network cost is compared with the bubble sort graph.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.11 no.5
• /
• pp.940-948
• /
• 2007

Many real world problems can be modeled as a graph and traversals on the graph. The structure of Web pages can be represented as a graph, for example, and user's navigation paths on the Web pages can be model as a traversal on the graph. It is interesting to discover valuable patterns, such as frequent patterns, from such traversals. In this paper, we propose an algorithm to discover frequent traversal patterns when a directed graph and weighted traversals on the graph are given. Furthermore, we propose a performance enhancement by traversal split and then verify it through experiments.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.14 no.4
• /
• pp.637-644
• /
• 2019

This paper describes an implementation of a graph-based simultaneous localization and mapping(SLAM) method called the General Graph Optimization. The General Graph Optimization formulates the SLAM problem using nodes and edges. The nodes represent the location and attitude of a robot in time sequence, and the edge between the nodes depict the constraint between the nodes. The constraints are imposed by sensor measurements. The General Graph Optimization solves the problem by optimizing the performance index determined by the constraints. The implementation is verified using the measurement data sets which are open for test of various SLAM methods.

• The Journal of the Acoustical Society of Korea
• /
• v.43 no.2
• /
• pp.234-242
• /
• 2024

We propose a passive sonar signal classification algorithm using Graph Neural Network (GNN). The proposed algorithm segments spectrograms into image patches and represents graphs through connections between adjacent image patches. Subsequently, Graph Convolutional Network (GCN) is trained using the represented graphs to classify signals. In experiments with publicly available underwater acoustic data, the proposed algorithm represents the line frequency features of spectrograms in graph form, achieving an impressive classification accuracy of 92.50 %. This result demonstrates a 8.15 % higher classification accuracy compared to conventional Convolutional Neural Network (CNN).

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.3
• /
• pp.91-100
• /
• 2020

Unlike the existing Visual Question Answering(VQA) problems, the new Visual Commonsense Reasoning(VCR) problems require deep common sense reasoning for answering questions: recognizing specific relationship between two objects in the image, presenting the rationale of the answer. In this paper, we propose a novel deep neural network model, KG_VCR, for VCR problems. In addition to make use of visual relations and contextual information between objects extracted from input data (images, natural language questions, and response lists), the KG_VCR also utilizes commonsense knowledge embedding extracted from an external knowledge base called ConceptNet. Specifically the proposed model employs a Graph Convolutional Neural Network(GCN) module to obtain commonsense knowledge embedding from the retrieved ConceptNet knowledge graph. By conducting a series of experiments with the VCR benchmark dataset, we show that the proposed KG_VCR model outperforms both the state of the art(SOTA) VQA model and the R2C VCR model.

• Journal of Convergence for Information Technology
• /
• v.10 no.8
• /
• pp.15-22
• /
• 2020

The 4th industrial revolution and the rapid change in the data environment revealed technical limitations in the existing relational database(RDB). As a new analysis method for unstructured data in all fields such as IDC/finance/insurance, interest in graph database(GDB) technology is increasing. The graph database is an efficient technique for expressing interlocked data and analyzing associations in a wide range of networks. This study extended the existing RDB to the GDB model and applied machine learning algorithms (pattern recognition, clustering, path distance, core extraction) to detect new abnormal signs. As a result of the performance analysis, it was confirmed that the performance of abnormal behavior(about 180 times or more) was greatly improved, and that it was possible to extract an abnormal symptom pattern after 5 steps that could not be analyzed by RDB.

• Journal of Internet Computing and Services
• /
• v.15 no.6
• /
• pp.125-132
• /
• 2014

Since frequent pattern mining was proposed in order to search for hidden, useful pattern information from large-scale databases, various types of mining approaches and applications have been researched. Especially, frequent graph pattern mining was suggested to effectively deal with recent data that have been complicated continually, and a variety of efficient graph mining algorithms have been studied. Graph patterns obtained from graph databases have their own importance and characteristics different from one another according to the elements composing them and their lengths. However, traditional frequent graph pattern mining approaches have the limitations that do not consider such problems. That is, the existing methods consider only one minimum support threshold regardless of the lengths of graph patterns extracted from their mining operations and do not use any of the patterns' weight factors; therefore, a large number of actually useless graph patterns may be generated. Small graph patterns with a few vertices and edges tend to be interesting when their weighted supports are relatively high, while large ones with many elements can be useful even if their weighted supports are relatively low. For this reason, we propose a weight-based frequent graph pattern mining algorithm considering length-decreasing support constraints. Comprehensive experimental results provided in this paper show that the proposed method guarantees more outstanding performance compared to a state-of-the-art graph mining algorithm in terms of pattern generation, runtime, and memory usage.