• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.05a
• /
• pp.394-397
• /
• 2018

최근 인공지능 연구가 활발하게 진행이 되면서 여러 기업에서 자율 주행연구도 활발하게 진행되고 있다. 하지만 실제 상황에서 자동차 주행 데이터를 얻기에는 여러 위험사항들과 경제적인 낭비가 있다. 그렇기 때문에 게임 상에서 데이터를 수집하고 딥러닝을 이용해 학습을 하기로 했다. 본 논문에서는 실제 세계와 유사한 환경을 가지고 있는 자동차 게임을 이용하여 자율 주행을 시도 했다. 자율 주행 시 많이 쓰이는 End to End 방법으로 데이터를 수집하면 두 가지 데이터가 저장된다. 하나는 이미지 데이터고 두 번째는 방향키 데이터다. 이러한 데이터들을 numpy 타입으로 40분간 데이터를 수집한 후 딥러닝에 많이 쓰이는 tensorflow를 사용하여 구현한 CNN을 이용하여 학습이 되는 것을 확인을 하고 91.9%의 정확도를 얻었다. 이를 기반으로 실세계에서의 사용 가능성을 확인했다.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2022.11a
• /
• pp.369-370
• /
• 2022

최근 자율 운항 선박에 대한 관심이 높아지고 있다. 특히, MUNIN (Maritime Unmanned Navigation through Intelligence in Networks) 프로젝트를 계기로 자율 운항 선박에 대한 개발과 연구가 활발히 진행되고 있다. 또한 국제해사기구 IMO는 자율 운항 선박 시대에 대응하기 위해 자율 선박을 MASS (Maritime Autonomous Surface Ship)라 정의하고 선박 자율화 정도에 따라 4단계 등급을 제시하고 있다. 완전한 자율 운항 선박에 대한 요구조건을 만족하기 위해서는 항로 결정과 제어기술이 필수적이다. 본 연구에서는 여러 가지 기술 중 선박의 최적경로를 생성하는 기법을 다룬다. 기존에 최적항로를 생성하기 위한 방법으로는 A^*, Dijkstra와 같은 알고리즘들이 주로 사용되었다. 그러나 이와 같은 알고리즘은 섬이나 육지에 대한 충돌 회피는 고려하고 있지만 수심 및 연안 선박에 대한 규정들은 고려하지 않고 있어 실제로 적용하기에는 한계점이 있다. 따라서 본 연구에서는 안전을 위해 선박의 선저 여유 수심과, 해도에 규정되어 있는 선박 운항에 대한 여러 규정들을 반영하여 최적 항로를 생성하고자 한다. 최적 항로를 생성하기 위한 알고리즘으로는 강화학습 기반의 Q-learning 알고리즘을 적용하였다.

• Journal of Gifted/Talented Education
• /
• v.23 no.6
• /
• pp.1077-1098
• /
• 2013

Science gifted students enrolled in a program, where classes had either explicit or implicit instruction about self-directed learning, were asked to write what was satisfying after each class. This process was part of the evaluation of the program. Students' descriptions related to self-directed learning are compared in these two classes, one with explicit instruction and the other with implicit instruction. First, most of the components related to self-directed learning, which were reported in the previous research articles, were mentioned in students evaluation. If there was any specific description regarding what was satisfying, there were components of self-directed learning. Students descriptions were consistent with list of self-directed learning components, which was constructed based on the previous research. Therefore it may be concluded that students recognized most of the reported self-directed learning components and satisfied with them. Second, There were differences in the evaluation of two types of classes. The evaluation of class with explicit instruction contained more self-directed learning components more frequently. For example, students worked in small groups in both classes. However more students mentioned small groups in classes with explicit instruction. As a result the explicit instruction appears to be more effective for students to recognize the self-directed learning components. However some of the components mentioned in classes with implicit instruction were not mentioned in the classes with explicit instruction. Therefore classes with explicit and implicit instructions are complimentary with each other and both instructions are necessary.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.38 no.1
• /
• pp.1-6
• /
• 2001

In this paper we reanalyze Kohonen‘s learning vector quantizing (LVQ) Learning rule which is based on Hcbb’s learning rule with a view to a gradient descent method. Kohonen's LVQ can be classified into two algorithms according to 6learning mode: unsupervised LVQ(ULVQ) and supervised LVQ(SLVQ). These two algorithms can be represented as gradient descent methods, if target values of output neurons are generated properly. As a result, we see that the LVQ learning method is a special case of a gradient descent method and also that LVQ is represented by a generalized percetron-like LVQ(PLVQ).

• The Journal of Information Technology
• /
• v.7 no.3
• /
• pp.11-18
• /
• 2004

Learning procedure in the neural network is updating of weights between neurons. Unadequate initial learning coefficient causes excessive iterations of learning process or incorrect learning results and degrades learning efficiency. In this paper, adaptive learning algorithm is proposed to increase the efficient in the learning algorithms of Self-Organization Neural Networks. The algorithm updates the weights adaptively when learning procedure runs. To prove the efficiency the algorithm is experimented to classification of strokes which is the reference handwritten character. The result shows improved classification rate about 1.44~3.65% proposed method compare with Kohonan and Mao's algorithms, in this paper.

• Convergence Security Journal
• /
• v.18 no.3
• /
• pp.3-9
• /
• 2018

The Deep Deterministic Policy Gradient (DDPG) algorithm is an algorithm that learns by using artificial neural network s and reinforcement learning. Among the studies related to reinforcement learning, which has been recently studied, the D DPG algorithm has an advantage of preventing the cases where the wrong actions are accumulated and affecting the learn ing because it is learned by the off-policy. In this study, we experimented to control the bicycle autonomously by applyin g the DDPG algorithm. Simulation was carried out by setting various environments and it was shown that the method us ed in the experiment works stably on the simulation.

• Journal of Internet of Things and Convergence
• /
• v.9 no.1
• /
• pp.65-70
• /
• 2023

In this paper, we propose an autonomous driving system using an end-to-end model to improve lane departure and misrecognition of traffic lights in a vision sensor-based system. End-to-end learning can be extended to a variety of environmental conditions. Driving data is collected using a model car based on a vision sensor. Using the collected data, it is composed of existing data and data with outlayers removed. A class was formed with camera image data as input data and speed and steering data as output data, and data learning was performed using an end-to-end model. The reliability of the trained model was verified. Apply the learned end-to-end model to the model car to predict the steering angle with image data. As a result of the learning of the model car, it can be seen that the model with the outlayer removed is improved than the existing model.

• Electronics and Telecommunications Trends
• /
• v.32 no.1
• /
• pp.25-34
• /
• 2017

무선통신, 이동통신 및 사물인터넷 기술의 발달에 힘입어 인터넷의 규모와 복잡도는 해마다 증가하고 있으며, 망의 제어와 관리의 복잡도 역시 함께 증가할 것으로 예상된다. 이에 따라 운용자(operator)가 담당하던 제어와 관리를 망이 스스로 수행하는 자율네트워킹(autonomic networking) 기술이 등장하게 되었다. 초기의 자율네트워킹 연구는 자가관리(self-management)를 위한 프레임워크를 개발하는 것에 중점을 두었으나, 이후에는 SDN/NFV 기반 플랫폼에 기계학습 기술을 접목함으로써, 유연성이 확보된 망에 지능화된 제어 및 관리를 제공하는 방향으로 진화하고 있다. 본고에서는 자율네트워킹에 관한 최근의 연구동향을 소개한다.

• The Journal of the Korea Contents Association
• /
• v.21 no.10
• /
• pp.594-602
• /
• 2021

This study aims to identify the learning behavior patterns recognized by students to find effective tutoring operational methods, and further analyze the impact of learning behavior patterns on academic performance and learning satisfaction. To this end, 105 participants in the tutoring program at D University based in Busan Metropolitan City collected data and conducted descriptive statistics, correlation analysis and regression analysis according to research problems. First, the study found that students who participated in tutoring had the most environment-dependent and self-taught learning behavioral styles and environment-independent and self-taught learning behavioral style. Second, the correlation between learning behavioral styles and academic achievement and learning satisfaction shows that there is a high correlation between positive and cooperative learning behavioral styles and environment-independent and self-taught learning behavioral styles. Third, regression analysis on academic achievement and learning satisfaction showed that positive and cooperative learning behavioral styles significantly affects learning satisfaction, but environment-independent and self-taught learning behavioral style, environment-dependent and self-taught learning behavioral style, and passive learning behavioral style were not significant. These results suggest that from the school perspective, learning behavior can be recognized as an important factor in students' academic success and failure, so instructors need to check learners' learning behavior patterns and provide appropriate tutoring teaching and learning design plans.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.515-517
• /
• 2021

원전 빅데이터를 효율적으로 분석하고 수집된 데이터를 인공지능 서비스에 활용할 수 있도록 제공하기 위해서는 원전 데이터에 특화된 빅데이터 플랫폼이 필요하다. 단순히 시간 순으로 나열된 원시(Raw) 데이터는 의미있는 단위로 논리적으로 구분되어 관리될 필요가 있고, 사건/사고의 발생에 따른 분류가 필요하다. 뿐만 아니라, 다수의 데이터들을 분석하여 수천 개의 계측신호들 중에서 원하는 목적에 적합한 신호가 어떠한 것들인지를 찾아낼 수 있는 데이터 분석이 지원될 필요가 있다. 이는 기계학습 애플리케이션을 개발할 때 필수적인 고품질의 데이터 제공에 크게 기여할 수 있다. 본 연구에서는 원전 데이터를 효과적으로 처리하고 분석하기 위한 원전 데이터 전처리 및 분석 기술을 고안하고 이를 빅데이터 저장 인프라와 통합한 원전 빅데이터 처리 체계를 소개한다. 본 연구의 결과물은 본격적인 원전 빅데이터 시스템 구축 사업에 활용될 것으로 기대된다.