• Journal of the Korea Society of Computer and Information
• /
• v.28 no.1
• /
• pp.27-38
• /
• 2023

In this paper, we propose a conversational AI agent based on continual learning that can continuously learn and grow with new data over time. A continual learning-based conversational AI agent consists of three main components: Task manager, User attribute extraction, and Auto-growing knowledge graph. When a task manager finds new data during a conversation with a user, it creates a new task with previously learned knowledge. The user attribute extraction model extracts the user's characteristics from the new task, and the auto-growing knowledge graph continuously learns the new external knowledge. Unlike the existing conversational AI agents that learned based on a limited dataset, our proposed method enables conversations based on continuous user attribute learning and knowledge learning. A conversational AI agent with continual learning technology can respond personally as conversations with users accumulate. And it can respond to new knowledge continuously. This paper validate the possibility of our proposed method through experiments on performance changes in dialogue generation models over time.

• Journal of KIISE
• /
• v.42 no.8
• /
• pp.998-1009
• /
• 2015

In recent years, there has been a growing interest in RDFS Inference to build a rich knowledge base. However, it is difficult to improve the inference performance with large data by using a single machine. Therefore, researchers are investigating the development of a RDFS inference engine for a distributed computing environment. However, the existing inference engines cannot process data in real-time, are difficult to implement, and are vulnerable to repetitive tasks. In order to overcome these problems, we propose a method to construct an in-memory distributed inference engine that uses a parallel graph structure. In general, the ontology based on a triple structure possesses a graph structure. Thus, it is intuitive to design a graph structure-based inference engine. Moreover, the RDFS inference rule can be implemented by utilizing the operator of the graph structure, and we can thus design the inference engine according to the graph structure, and not the structure of the data table. In this study, we evaluate the proposed inference engine by using the LUBM1000 and LUBM3000 data to test the speed of the inference. The results of our experiment indicate that the proposed in-memory distributed inference engine achieved a performance of about 10 times faster than an in-storage inference engine.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1939-1944
• /
• 2003

This paper presents a global path planning algorithm based on a visibility graph method, and applies additionally various constraints for constructing the reduced visibility graph. The modification algorithm for generating the rounded path is applied to the globally shortest path of the visibility graph using the robot size constraint in order to avoid the obstacle. In order to check the visibility in given 3D map data, 3D CAD data with VRML format is projected to the 2D plane of the mobile robot, and the projected map is converted into an image for easy map analysis. The image processing are applied to this grid map for extracting the obstacles and the free space. Generally, the tree size of visibility graph is proportional to the factorial of the number of the corner points. In order to reduce the tree size and search the shortest path efficiently, the various constraints are proposed. After short paths that crosses the corner points of obstacles lists up, the shortest path among these paths is selected and it is modified to the combination of the line path and the arc path for the mobile robot to avoid the obstacles and follow the rounded path in the environment. The proposed path planning algorithm is applied to the mobile robot LCAR-III.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.9
• /
• pp.3598-3614
• /
• 2020

With the increase of motor vehicles and tourism demand, some traffic problems gradually appear, such as traffic congestion, safety accidents and insufficient allocation of traffic resources. Facing these challenges, a model of Spatio-Temporal Dilated Convolutional Network (STDGCN) is proposed for assistance of extracting highly nonlinear and complex characteristics to accurately predict the future traffic flow. In particular, we model the traffic as undirected graphs, on which graph convolutions are built to extract spatial feature informations. Furthermore, a dilated convolution is deployed into graph convolution for capturing multi-scale contextual messages. The proposed STDGCN integrates the dilated convolution into the graph convolution, which realizes the extraction of the spatial and temporal characteristics of traffic flow data, as well as features of road occupancy. To observe the performance of the proposed model, we compare with it with four rivals. We also employ four indicators for evaluation. The experimental results show STDGCN's effectiveness. The prediction accuracy is improved by 17% in comparison with the traditional prediction methods on various real-world traffic datasets.

• The Journal of Society for e-Business Studies
• /
• v.20 no.2
• /
• pp.155-166
• /
• 2015

RWR (Random Walk with Restart) is frequently used by many graph-based ranking algorithms, but it does not consider a signed graph where edges may have negative weight values. In this paper, we apply the Balance Theory by F. Heider to RWR over a signed graph and propose a novel RWR, Balanced Random Walk (BRW). We apply the proposed technique into the domain of recommendation system, and show by experiments its effectiveness to filter out the items that users may dislike. In order to provide the reasonable performance of BRW in the domain, we modify the existing Top-k algorithm, BCA, and propose a new algorithm, Bicolor-BCA. The proposed algorithm yet requires employing a threshold. In the experiment, we show how threshold values affect both precision and performance of the algorithm.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.20 no.2
• /
• pp.234-242
• /
• 2010

In this paper, we deal with the visualization of graphs, which are one of the most important data structures. As the size of a graph increases, it becomes more difficult to check the graph visually because of the increase of edge intersections. We propose a new method of overcoming such problem. Most of previous studies considered only the minimization of edge intersections, but we additionally pursue to preserve distances between vertices. We present a novel genetic algorithm using an evaluation function based on a weighted sum of two objectives. Our experiments could show effective visualization results.

• Proceedings of the KSRS Conference
• /
• 2002.10a
• /
• pp.524-529
• /
• 2002

In this paper, we report the development of two DEM (digital elevation model) generation algorithms over urban areas from an IKONOS stereo pair. One ("EpiMatch") is originally developed for SPOT images and modified for IKONOS images. It uses epipolar geometry for accurate DEM generation. The other is based on graph-cut algorithm in 3D voxel space. This algorithm is believed to work better on height discontinuities than EpiMatch. An IKONOS image pair over Taejon city area was used for tests. Using ground control points obtained from differential GPS, camera model was set up and stereo matching applied. As a result, two DEMs over urban areas were produced. Within a DEM from EpiMatch small houses appear as small "cloudy" patches and large apartment and industrial buildings are visually identifiable. Within the DEM from graph-cut we could achieve better height information on building boundaries. The results show that both algorithms can generate DEMs from IKONOS images although more research is required on handling height discontinuities (for "EpiMatch") and on faster computation (for "Graph-cut").

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.27 no.4
• /
• pp.324-341
• /
• 2023

Topological data analysis (TDA) is a data analysis technique, recently developed, that investigates the overall shape of a given dataset. The mapper algorithm is a TDA method that considers the connectivity of the given data and converts the data into a mapper graph. Compared to persistent homology, another popular TDA tool, that mainly focuses on the homological structure of the given data, the mapper algorithm is more of a visualization method that represents the given data as a graph in a lower dimension. As it visualizes the overall data connectivity, it could be used as a prediction method that visualizes the new input points on the mapper graph. The existing mapper packages such as Giotto-TDA, Gudhi and Kepler Mapper provide the descriptive mapper algorithm, that is, the final output of those packages is mainly the mapper graph. In this paper, we develop a simple predictive algorithm. That is, the proposed algorithm identifies the node information within the established mapper graph associated with the new emerging data point. By checking the feature of the detected nodes, such as the anomality of the identified nodes, we can determine the feature of the new input data point. As an example, we employ the fraud credit card transaction data and provide an example that shows how the developed algorithm can be used as a node prediction method.

• Journal of Computing Science and Engineering
• /
• v.10 no.3
• /
• pp.85-94
• /
• 2016

We propose a model-based method for information alignment using educational standards as a case study. Discrepancies and inconsistencies in educational standards across different states/cities hinder the retrieval and sharing of educational resources. Unlike existing educational standards alignment systems that only give binary judgments (either "aligned" or "not-aligned"), our proposed system classifies each pair of educational standard statements in one of seven levels of alignments: Strongly Fully-aligned, Weakly Fully-aligned, Partially-$aligned^{***}$, Partially-$aligned^{**}$, Partially-$aligned^*$, Poorly-aligned, and Not-aligned. Such a 7-level categorization extends the notion of binary alignment and provides a finer-grained system for comparing educational standards that can broaden categories of resource discovery and retrieval. This study continues our previous use of mathematics education as a domain, because of its generally unambiguous concepts. We adopt a materialization pattern (MP) model developed in our earlier work to represent each standard statement as a verb-phrase graph and a noun-phrase graph; we align a pair of statements using graph matching based on Bloom's Taxonomy, WordNet, and taxonomy of mathematics concepts. Our experiments on data sets of mathematics educational standards show that our proposed system can provide alignment results with a high degree of agreement with domain expert's judgments.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.5
• /
• pp.1463-1478
• /
• 2012

Based on the theory of local-world network, the composition self-similarity (CSS) of network traffic is presented for the first time in this paper for the study of DoS detection. We propose the concept of composition distribution graph and design the relative operations. The $(R/S)^d$ algorithm is designed for calculating the Hurst parameter. Based on composition distribution graph and Kullback Leibler (KL) divergence, we propose the composition self-similarity anomaly detection (CSSD) method for the detection of DoS attacks. We evaluate the effectiveness of the proposed method. Compared to other entropy based anomaly detection methods, our method is more accurate and with higher sensitivity in the detection of DoS attacks.