• The Journal of the Korea Contents Association
• /
• v.22 no.1
• /
• pp.35-46
• /
• 2022

Recently, with the development of network technologies, as IoT and social network service applications have been actively used, a lot of graph stream data is being generated. In this paper, we propose a graph compression scheme that considers the stream graph environment by applying graph mining to the existing compression technique, which has been focused on compression rate and runtime. In this paper, we proposed Incremental frequent pattern based compression technique for graph streams. Since the proposed scheme keeps only the latest reference patterns, it increases the storage utilization and improves the query processing time. In order to show the superiority of the proposed scheme, various performance evaluations are performed in terms of compression rate and processing time compared to the existing method. The proposed scheme is faster than existing similar scheme when the number of duplicated data is large.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.195-206
• /
• 2022

Crack detection is essential for inspection of existing structures and crack segmentation based on deep learning is a significant solution. However, datasets are usually one of the key issues. When building a new dataset for deep learning, laborious and time-consuming annotation of a large number of crack images is an obstacle. The aim of this study is to develop an approach that can automatically select a small portion of the most informative crack images from a large pool in order to annotate them, not to label all crack images. An active learning method with difficulty learning mechanism for crack segmentation tasks is proposed. Experiments are carried out on a crack image dataset of a steel box girder, which contains 500 images of 320×320 size for training, 100 for validation, and 190 for testing. In active learning experiments, the 500 images for training are acted as unlabeled image. The acquisition function in our method is compared with traditional acquisition functions, i.e., Query-By-Committee (QBC), Entropy, and Core-set. Further, comparisons are made on four common segmentation networks: U-Net, DeepLabV3, Feature Pyramid Network (FPN), and PSPNet. The results show that when training occurs with 200 (40%) of the most informative crack images that are selected by our method, the four segmentation networks can achieve 92%-95% of the obtained performance when training takes place with 500 (100%) crack images. The acquisition function in our method shows more accurate measurements of informativeness for unlabeled crack images compared to the four traditional acquisition functions at most active learning stages. Our method can select the most informative images for annotation from many unlabeled crack images automatically and accurately. Additionally, the dataset built after selecting 40% of all crack images can support crack segmentation networks that perform more than 92% when all the images are used.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.6
• /
• pp.1081-1090
• /
• 2022

The face recognition model is used for identity recognition of smartphones, providing convenience to many users. As a result, the security review of the DNN model is becoming important, with adversarial attacks present as a well-known vulnerability of the DNN model. Adversarial attacks have evolved to decision-based attack techniques that use only the recognition results of deep learning models to perform attacks. However, existing decision-based attack technique[14] have a problem that requires a large number of queries when generating adversarial examples. In particular, it takes a large number of queries to approximate the gradient. Therefore, in this paper, we propose a method of generating adversarial examples using orthogonal space sampling and dimensionality reduction sampling to avoid wasting queries that are consumed to approximate the gradient of existing decision-based attack technique[14]. Experiments show that our method can reduce the perturbation size of adversarial examples by about 2.4 compared to existing attack technique[14] and increase the attack success rate by 14% compared to existing attack technique[14]. Experimental results demonstrate that the adversarial example generation method proposed in this paper has superior attack performance.

• The Journal of the Korea Contents Association
• /
• v.22 no.8
• /
• pp.33-43
• /
• 2022

Labeled graphs are used to represent entities, their relationships, and their structures in real data such as knowledge graphs and protein interactions. With the rapid development of IT and the explosive increase in data, there has been a need for a subgraph matching technology to provide information that the user is interested in. In this paper, we propose an approximate Top-k labeled subgraph matching scheme that considers the semantic similarity of labels and the difference in graph structure. The proposed scheme utilizes a learning model using FastText in order to consider the semantic similarity of a label. In addition, the label similarity graph(LSG) is used for approximate subgraph matching by calculating similarity values between labels in advance. Through the LSG, we can resolve the limitations of the existing schemes that subgraph expansion is possible only if the labels match exactly. It supports structural similarity for a query graph by performing searches up to 2-hop. Based on the similarity value, we provide k subgraph matching results. We conduct various performance evaluations in order to show the superiority of the proposed scheme.

• Bioinformatics and Biosystems
• /
• v.1 no.2
• /
• pp.99-102
• /
• 2006

The Basic Local Alignment Search Tool (BLAST) is one of the most established software in bioinformatics research and it compares a query sequence against the libraries of known sequences in order to investigate sequence similarity. Expressed Sequence Tags (ESTs) are single-pass sequence reads from mRNA (or cDNA) and represent the expression for a given cDNA library and the snapshot of genes expressed in a given tissue and/or at a given developmental stage. Therefore, ESTs can be very valuable information for functional genomics and bioinformatics researches. Although major bio database (DB) websites including NCBI are providing BLAST services and EST data, local DB and search system is demanding for better performance and security issue. Here we present animal EST DBs and local BLAST search system. The animal ESTs DB in NCBI Genbank were divided by animal species using the Perl script we developed. and we also built the new extended DB search systems fur the new data (Local Animal BLAST Search System: http://bioinfo.kohost.net), which was constructed on the high-capacity PC Cluster system fur the best performance. The new local DB contains 650,046 sequences for Bos taurus(cattle), 368,120 sequences for Sus scrofa (pig), 693,005 sequences for Gallus gallus (fowl), respectively.

• Journal of Digital Convergence
• /
• v.16 no.9
• /
• pp.147-157
• /
• 2018

Since Google launched a prediction service for influenza-like illness(ILI), studies on ILI prediction based on web search data have proliferated worldwide. In this regard, this study aims to build short-term predictive models for ILI in Korea using ILI and web search data and measure the performance of the said models. In these proposed ILI predictive models specific to Korea, ILI surveillance data of Korea CDC and Korean web search data of Google and Naver were used along with the ARIMA model. Model 1 used only ILI data. Models 2 and 3 added Google and Naver search data to the data of Model 1, respectively. Model 4 included a common query used in Models 2 and 3 in addition to the data used in Model 1. In the training period, the goodness of fit of all predictive models was higher than 95% ($R^2$). In predictive periods 1 and 2, Model 1 yielded the best predictions (99.98% and 96.94%, respectively). Models 3(a), 4(b), and 4(c) achieved stable predictability higher than 90% in all predictive periods, but their performances were not better than that of Model 1. The proposed models that yielded accurate and stable predictions can be applied to early warning systems for the influenza pandemic in Korea, with supplementary studies on improving their performance.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.3
• /
• pp.33-41
• /
• 2007

Owing to the development of multimedia and communication technologies, multimedia data become a common feature of the information systems and are on the increase. This has led to the need of 3D shape retrieval systems that, given a query object, retrieve similar 3D objects. Therefore, shape descriptor required to describe a 3D object effectively and efficiently. In this paper, a new descriptor for 3D model retrieval based on shape information is proposed. The proposed descriptor utilizes the curvedness together with the shape index that provides local geometry information. The existing 3D Shape Spectrum Descriptor (3D SSD), which is defined as the histogram of shape index values, represents the characteristics of local shapes of the 3D surface. However, it does not properly represent the local shape characteristics, because many points with different curvedness may have the same shape index value. Therefore, we add a new feature that represents the degree of curvedness, thereby improving the discriminating power of the shape descriptor. We evaluate the performance of the proposed method, compared with the previous method. The experimental results have shown that the performance of retrieval has been improved by 23.6%.

• Journal of Korea Spatial Information System Society
• /
• v.7 no.1 s.13
• /
• pp.63-76
• /
• 2005

Recently, the wireless positioning techniques and mobile computing techniques were developed with rapidly to use location data of moving objects. Also, the demand for LBS(Location Based Services) which uses location data of moving objects is increasing rapidly. In order to support various LBS, a system that can store and retrieve location data of moving objects efficiently is required necessarily. The more the number of moving objects is numerous and the more periodical sampling of locations is frequent, the more location data of moving objects become very large. Hence the system should be able to efficiently manage mass location data, support various spatio-temporal queries for LBS, and solve the uncertainty problem of moving objects. Therefore, in this paper, we presented a hash technique, a clustering technique and a trajectory search technique to manage location data of moving objects efficiently And, we have developed a Mass Moving Object Location Data Management System, which is a disk-based system, that can store and retrieve location data of mass moving objects efficiently and support the query for spatio-temporal data and the past location data with uncertainty. By analying the performance of the Mass Moving Object Locations Management system and the SQL-Server, we can find that the performance of our system for storing and retrieving location data of moving objects was about 5% and 300% better than the SQL-Server, repectively.

• Journal of Korea Spatial Information System Society
• /
• v.11 no.1
• /
• pp.55-62
• /
• 2009

In u-GIS environment, we collect spatial data needed through sensor network and provide them with information real-time processed or stored. When information through Internet is requested on Web based applications, it is transmitted in XML. Especially, when requested information includes spatial data, GML, S-XML, and other document that can process spatial data are used. In this processing, real-time stream data processed in DSMS is transformed to S-XML document type and spatial information service based on web receive S-XML document through Internet. Because most of spatial application service use existing spatial DBMS as a storage system, The data used in S-XML and SDBMS needs transformation between themselves. In this paper, we propose S-XML a transformation method using caching of spatial data. The proposed method caches the spatial data part of S-XML to transform S-XML and relational spatial database for providing spatial data efficiently and it transforms cached data without additional transformation cost when a transformation between data in the same region is required. Through proposed method, we show that it reduced the cost of transformation between S-XML documents and spatial information services based on web to provide spatial information in u-GIS environment and increased the performance of query processing through performance evaluation.

• Journal of KIISE
• /
• v.44 no.8
• /
• pp.813-821
• /
• 2017

Entity Linking find the meaning of an entity mention, which indicate the entity using different expressions, in a user's query by linking the entity mention and the entity in the knowledge base. This task has four challenges, including the difficult knowledge base construction problem, multiple presentation of the entity mention, ambiguity of entity linking, and NIL entity recognition. In this paper, we first construct the entity name dictionary based on Wikipedia to build a knowledge base and solve the multiple presentation problem. We then propose various methods for NIL entity recognition and solve the ambiguity of entity linking by training the support vector machine based on several features, including the similarity of the context, semantic relevance, clue word score, named entity type similarity of the mansion, entity name matching score, and object popularity score. We sequentially use the proposed two methods based on the constructed knowledge base, to obtain the good performance in the entity linking. In the result of the experiment, our system achieved 83.66% and 90.81% F1 score, which is the performance of the NIL entity recognition to solve the ambiguity of the entity linking.