• Proceedings of the Korea Information Processing Society Conference
• /
• 2002.04b
• /
• pp.805-808
• /
• 2002

컴퓨터 시스템 및 네트워크에 대한 침입 공격의 방법 중 이미 알려진 형태의 공격에 대해서는 상대적으로 탐지가 용이하나 사용자의 비정상행위는 방법의 다양성 때문에 탐지가 매우 어렵다. 그러나, 사용자의 정상적인 행동은 몇 가지 소수의 형태로 특정 지어질 수 있다. 본 논문에서는 상대적으로 변화가 적은 정상 행위를 신경망으로 Modeling하여 이를 비정상 행위 탐지에 적용하는 기법을 제안한다. 이를 위하여 입력 영역을 지역화 하는 특성을 갖는 RBF(Radial-Basis-Fuction) 신경망에 대한 단일 Class의 학습방법을 제안하고, 이를 이용한 비정상 행위에 대한 공격의 탐지에 대한 적용 방안을 제시한다. 비정상 행위 탐지에 대한 적용 가능성을 검증하기 위하여 사용자가 키보드 입력 유형을 학습하고 이를 이용하여 타인의 ID와 Password를 도용한 경우의 탐지에 적용하였다.

• Proceedings of the Korean Institute of Surface Engineering Conference
• /
• 2009.10a
• /
• pp.237-238
• /
• 2009

본 연구에서는 이온에너지와 박막 표면 거칠기와의 관계를 신경망을 이용하여 모델링하였다. Pulsed 플라즈마 증착장비를 이용하여 상온에서 실리콘 나이트라이드 (SiN)을 증착하였다. 바이어스 전력과 duty ratio는 각각 $40{\sim}100W$와 $30{\sim}90%$로 변화하였다. 이온에너지 정보는 비침투식 이온에너지 분석시스템을 이용하여 수집하였다. 신경망의 성능은 유전자알고리즘을 이용하여 최적화시켰다. 최적화한 모델은 이온에너지의 영향을 고찰하였다. 모델로부터 고 이온 에너지는 저 이온에너지가 높은 조건에서 증가시킬 때에 표면 거칠기를 보다 작게 한다는 것을 알 수 있었다.

• ETRI Journal
• /
• v.43 no.2
• /
• pp.288-298
• /
• 2021

Abstractive text summarization is a process of making a summary of a given text by paraphrasing the facts of the text while keeping the meaning intact. The manmade summary generation process is laborious and time-consuming. We present here a summary generation model that is based on multilayered attentional peephole convolutional long short-term memory (MAPCoL; LSTM) in order to extract abstractive summaries of large text in an automated manner. We added the concept of attention in a peephole convolutional LSTM to improve the overall quality of a summary by giving weights to important parts of the source text during training. We evaluated the performance with regard to semantic coherence of our MAPCoL model over a popular dataset named CNN/Daily Mail, and found that MAPCoL outperformed other traditional LSTM-based models. We found improvements in the performance of MAPCoL in different internal settings when compared to state-of-the-art models of abstractive text summarization.

• Journal of Information Processing Systems
• /
• v.15 no.1
• /
• pp.91-99
• /
• 2019

An innovative tone modeling framework based on deep neural networks in tone recognition was proposed in this paper. In the framework, both the prosodic features and the articulatory features were firstly extracted as the raw input data. Then, a 5-layer-deep deep belief network was presented to obtain high-level tone features. Finally, support vector machine was trained to recognize tones. The 863-data corpus had been applied in experiments, and the results show that the proposed method helped improve the recognition accuracy significantly for all tone patterns. Meanwhile, the average tone recognition rate reached 83.03%, which is 8.61% higher than that of the original method.

• Structural Engineering and Mechanics
• /
• v.85 no.1
• /
• pp.15-27
• /
• 2023

Using artificial intelligence and internet of things methods in engineering and industrial problems has become a widespread method in recent years. The low computational costs and high accuracy without the need to engage human resources in comparison to engineering demands are the main advantages of artificial intelligence. In the present paper, a deep neural network (DNN) with a specific method of optimization is utilize to predict fundamental natural frequency of a cylindrical structure. To provide data for training the DNN, a detailed numerical analysis is presented with the aid of functionally modified couple stress theory (FMCS) and first-order shear deformation theory (FSDT). The governing equations obtained using Hamilton's principle, are further solved engaging generalized differential quadrature method. The results of the numerical solution are utilized to train and test the DNN model. The results are validated at the first step and a comprehensive parametric results are presented thereafter. The results show the high accuracy of the DNN results and effects of different geometrical, modeling and material parameters in the natural frequencies of the structure.

• International Journal of Advanced Culture Technology
• /
• v.11 no.3
• /
• pp.190-198
• /
• 2023

Sentiment analysis can be applied to all texts generated from websites, blogs, messengers, etc. The study fulfills an artificial intelligence sentiment analysis estimating visiting evaluation opinions (reviews) and visitor ratings, and suggests a deep learning model which foretells either an affirmative or a negative inclination for new reviews. This study operates review big data about Jeju tourist attractions which are extracted from Google from October 1st, 2021 to November 30th, 2021. The normalization data used in the propensity prediction modeling of this study were divided into training data and test data at a 7.5:2.5 ratio, and the CNN classification neural network was used for learning. The predictive model of the research indicates an accuracy of approximately 84.72%, which shows that it can upgrade performance in the future as evaluating its error rate and learning precision.

• Advances in aircraft and spacecraft science
• /
• v.10 no.6
• /
• pp.521-543
• /
• 2023

Reynolds-averaged Navier-Stokes (RANS) models are extensively employed in industrial settings for the purpose of simulating intricate fluid flows. However, these models are subject to certain limitations. Notably, disparities persist in the Reynolds stresses when comparing the RANS model with high-fidelity data obtained from Direct Numerical Simulation (DNS) or experimental measurements. In this work we propose an approach to mitigate these discrepancies while retaining the favorable attributes of the Menter Shear Stress Transport (SST) model, such as its significantly lower computational expense compared to DNS simulations. This strategy entails incorporating an explicit algebraic model and employing a neural network to correct the turbulent characteristic time. The imposition of realizability constraints is investigated through the introduction of penalization terms. The assimilated Reynolds stress model demonstrates good predictive performance in both in-sample and out-of-sample flow configurations. This suggests that the model can effectively capture the turbulent characteristics of the flow and produce physically realistic predictions.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.05a
• /
• pp.476-478
• /
• 2018

현대 사회의 발전으로 인해 사람들의 삶의 질이 향상됨에 따라 사람들은 다양한 방식으로 자신 및 자신의 개성을 표출하려는 시도를 한다. 특히 IT 기술의 발전은 가상현실 및 3D 기술의 성장을 이끌어냈다. 본 논문은 다가올 4차 산업혁명에 발맞추어 사용자의 개성을 표출할 실용적이고 개성 있는 3D 모델링 아이디어를 제안하고자 한다. 스마트폰 사진 촬영과 동시에 사용자가 선택한 다른 캐릭터 사진과의 합성 사진을 Convolutional Neural Network (CNN)과 Generative Adversarial Network (GAN) 기반 딥러닝 기술을 통해 생성한다. 생성된 이미지는 사용자의 모습과 합성의 대상이 되는 캐릭터의 모습을 동시에 담고 있다. 본 연구의 결과물로 생성된 합성 사진을 3D 프린터를 이용하여 자신만의 모습이 담긴 굿즈를 생산 혹은 이모티콘을 생성하는 등 다양한 실용적인 응용분야에 적용 가능하다.

• Journal of Information Processing Systems
• /
• v.17 no.4
• /
• pp.787-800
• /
• 2021

Three-dimensional (3D) object classification tasks using point clouds are widely used in 3D modeling, face recognition, and robotic missions. However, processing raw point clouds directly is problematic for a traditional convolutional network due to the irregular data format of point clouds. This paper proposes a pointwise convolution neural network (CNN) structure that can process point cloud data directly without preprocessing. First, a 2D convolutional layer is introduced to percept coordinate information of each point. Then, multiple 2D convolutional layers and a global max pooling layer are applied to extract global features. Finally, based on the extracted features, fully connected layers predict the class labels of objects. We evaluated the proposed pointwise CNN structure on the ModelNet10 dataset. The proposed structure obtained higher accuracy compared to the existing methods. Experiments using the ModelNet10 dataset also prove that the difference in the point number of point clouds does not significantly influence on the proposed pointwise CNN structure.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.7
• /
• pp.2321-2338
• /
• 2021

Compared with vehicle trajectories, pedestrian trajectories have stronger degrees of freedom and complexity, which poses a higher challenge to trajectory prediction tasks. This paper designs a mode to divide the trajectory of pedestrians at a traffic intersection, which converts the trajectory regression problem into a trajectory classification problem. This paper builds a deep model for pedestrian trajectory prediction at intersections for the task of pedestrian short-term trajectory prediction. The model calculates the spatial correlation and temporal dependence of the trajectory. More importantly, it captures the interactive features among pedestrians through the Attention mechanism. In order to improve the training speed, the model is composed of pure convolutional networks. This design overcomes the single-step calculation mode of the traditional recurrent neural network. The experiment uses Vulnerable Road Users trajectory dataset for related modeling and evaluation work. Compared with the existing models of pedestrian trajectory prediction, the model proposed in this paper has advantages in terms of evaluation indicators, training speed and the number of model parameters.