• Geophysics and Geophysical Exploration
• /
• v.24 no.3
• /
• pp.89-97
• /
• 2021

Fault detection in seismic data is well suited to the application of machine learning algorithms. Accordingly, various machine learning techniques are being developed. In recent studies, machine learning models, which utilize synthetic data, are the particular focus when training with deep learning. The use of synthetic training data has many advantages; Securing massive data for training becomes easy and generating exact fault labels is possible with the help of synthetic training data. To interpret real data with the model trained by synthetic data, the synthetic data used for training should be geologically realistic. In this study, we introduce a method to generate realistic synthetic seismic data. Initially, reflectivity models are generated to include realistic fault structures, and then, a one-way wave equation is applied to efficiently generate seismic stack sections. Next, a migration algorithm is used to remove diffraction artifacts and random noise is added to mimic actual field data. A convolutional neural network model based on the U-Net structure is used to verify the generated synthetic data set. From the results of the experiment, we confirm that realistic synthetic data effectively creates a deep learning model that can be applied to field data.

• Journal of Practical Engineering Education
• /
• v.15 no.1
• /
• pp.223-232
• /
• 2023

The Work-Study parallel program has been promoted as a key policy to resolve the mismatch between industrial sites and school education and realize a competency-centered society, and as of December 2022, 16,664 companies participated in the training. Learning companies play a very important role as education and training supply organizations that conduct field training. In this study, for the evaluation of learning companies participating in work-study program, the authors derive important factors that determine the quality of on-site education and training by analyzing the cognitive structure of experts in charge of the company and present evaluation indicators for learning enterprises. Therefore, it is intended to lay the foundation for promoting the quality of work-study parallel program.

• Journal of the Korean Society for information Management
• /
• v.19 no.4
• /
• pp.35-51
• /
• 2002

Inductive learning and classification techniques have been employed in various research and applications that organize textual data to solve the problem of information access. In this study, we develop hybrid model combination methods which incorporate the concepts and techniques for multiple modeling algorithms to improve the accuracy of text classification, and conduct experiments to evaluate the performances of proposed schemes. Boosted stacking, one of the extended stacking schemes proposed in this study yields higher accuracy relative to the conventional model combination methods and single classifiers.

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.04b
• /
• pp.646-648
• /
• 2004

기계학습의 군집화(clustering) 기법은 예제들 간의 유사성에 근거하여 주어진 예제들을 무리 짓는 방법이다. 준감독(semi-supervised) 군집화는 카테고리가 부여된(labeled) 소수의 예제들을 적극적으로 활용하여 군집형태가 보다 자연스럽게 형성되도록 유도하는 군집화 방법이다. 준감독 군집화 문제에서 예제에 카테고리를 부여하는 작업은 현실적으로 극히 제한적이거나 카테고리를 부여하는데 소요되는 비용이 상당하므로, 제한된 자원 내에서 군집화에 효용성이 높을 예제들을 선정하여 카테고리를 부여하는 것이 필요하다. 본 논문에서는 기존 연구에서 능동적 학습의 초기 훈련예제 선정을 위해 제안된 군집기반 훈련예제 선정 방법을 준감독 군집화에 적용하여 군집 결과의 질을 향상시키고자 한다. 군집화를 이용한 예제 선정 방법은 유사한 예제들은 동일한 카테고리에 속할 가능성이 높다는 가정하에 전체 예제를 활용하여 선정하고자 하는 예제 수만큼 군집을 생성 한 후. 각 군집의 중심점에 가장 가까운 예제들을 대표 예제로 선정하여 훈련 집합을 구성하는 방법이다 본 논문에서는 문서를 대상으로 하는 준감독 군집화 실험을 통해, 카테고리를 부여할 예제를 임의로 선정한 경우에 비해 군집화를 이용한 훈련 예제들로 준감독 군집화를 수행한 경우가 보다 좋은 군집을 형성함을 확인하였다.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.324-328
• /
• 2022

기계 독해란 주어진 문서를 이해하고 문서 내의 내용에 대한 질문에 답을 추론하는 연구 분야이며, 기계 독해 문제의 종류 중에는 여러 개의 선택지에서 질문에 대한 답을 선택하는 객관식 형태의 문제가 존재한다. 이러한 자연어 처리 문제를 해결하기 위해 기존 연구에서는 사전학습된 언어 모델을 미세조정하여 사용하는 방법이 널리 활용되고 있으나, 학습 데이터가 부족한 환경에서는 기존의 일반적인 미세조정 방법으로 모델의 성능을 높이는 것이 제한적이며 사전학습된 의미론적인 정보를 충분히 활용하지 못하여 성능 향상에 한계가 있다. 이에 본 연구에서는 기존의 일반적인 미세조정 방법에 템플릿을 적용한 템플릿 기반 미세조정 방법을 통해 사전학습된 의미론적인 정보를 더욱 활용할 수 있도록 한다. 객관식 형태의 기계 독해 문제 중 하나인 토익 문제에 대해 모델을 템플릿 기반 미세조정 방법으로 실험을 진행하여 템플릿이 모델 학습에 어떠한 영향을 주는지 확인하였다.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.6
• /
• pp.187-194
• /
• 2020

This paper proposes a website fingerprinting method using ensemble learning over a Tor network that guarantees client anonymity and personal information. We construct a training problem for website fingerprinting from the traffic packets collected in the Tor network, and compare the performance of the website fingerprinting system using tree-based ensemble models. A training feature vector is prepared from the general information, burst, cell sequence length, and cell order that are extracted from the traffic sequence, and the features of each website are represented with a fixed length. For experimental evaluation, we define four learning problems (Wang14, BW, CW_T, CW_H) according to the use of website fingerprinting, and compare the performance with the support vector machine model using CUMUL feature vectors. In the experimental evaluation, the proposed statistical-based training feature representation is superior to the CUMUL feature representation except for the BW case.

• Geophysics and Geophysical Exploration
• /
• v.26 no.3
• /
• pp.114-125
• /
• 2023

The physics-informed neural network (PINN) has been proposed to overcome the limitations of various numerical methods used to solve partial differential equations (PDEs) and the drawbacks of purely data-driven machine learning. The PINN directly applies PDEs to the construction of the loss function, introducing physical constraints to machine learning training. This technique can also be applied to wave equation modeling. However, to solve the wave equation using the PINN, second-order differentiations with respect to input data must be performed during neural network training, and the resulting wavefields contain complex dynamical phenomena, requiring careful strategies. This tutorial elucidates the fundamental concepts of the PINN and discusses considerations for wave equation modeling using the PINN approach. These considerations include spatial coordinate normalization, the selection of activation functions, and strategies for incorporating physics loss. Our experimental results demonstrated that normalizing the spatial coordinates of the training data leads to a more accurate reflection of initial conditions in neural network training for wave equation modeling. Furthermore, the characteristics of various functions were compared to select an appropriate activation function for wavefield prediction using neural networks. These comparisons focused on their differentiation with respect to input data and their convergence properties. Finally, the results of two scenarios for incorporating physics loss into the loss function during neural network training were compared. Through numerical experiments, a curriculum-based learning strategy, applying physics loss after the initial training steps, was more effective than utilizing physics loss from the early training steps. In addition, the effectiveness of the PINN technique was confirmed by comparing these results with those of training without any use of physics loss.

• Review of KIISC
• /
• v.27 no.3
• /
• pp.27-32
• /
• 2017

머신러닝을 위주로 하는 인공지능 기술이 여러 분야에서 다양하게 적용되고 있다. 머신러닝 기술은 시험 데이터에 대해 높은 성능을 보였지만, 악의적으로 만들어진 데이터에 대해서는 오동작을 하는 경우가 보고되고 있다. 그 외에도 학습데이터 오염시키기, 학습된 모델 탈취 등 새로운 공격 유형이 보고되고 있다. 기계학습에 사용된 훈련데이터에 대한 보안과 프라이버시 또한 중요한 이슈이다. 인공지능 기술의 개발 및 적용에 있어 이러한 위험성에 대한 고려와 대비가 반드시 필요하다.

• Annual Conference of KIPS
• /
• 2021.11a
• /
• pp.914-917
• /
• 2021

IoT 기기는 취약한 아이디와 비밀번호 사용, 저사양 하드웨어 등 보안 취약점으로 인해 사이버 공격 진입점으로 이용되고 있다. 본 논문은 IoT 악성코드를 탐지하기 위한 op 코드 카테고리 기반 특징 표현을 제안한다. Op 코드의 기능별 분류 정보를 이용해서 n-gram 특징과 엔트로피 히스토그램 특징을 추출하고 IoT 악성코드 탐지를 위한 기계학습 모델 평가를 수행한다. IoT 악성코드는 기능 개선과 추가를 통해 진화하였으나 기계학습 모델은 훈련 데이터에 포함되지 않은 진화된 IoT 악성 코드에 대한 예측 성능이 우수하였다. 또한 특징 시각화를 이용해서 악성코드의 비교 탐지가 가능하다.

• Journal of the Korea Society for Simulation
• /
• v.23 no.2
• /
• pp.73-84
• /
• 2014

Recently, researches on the application of IT technology to various fields including traditional industries are becoming more popular. One challenge in the field of education is to understand the way how technology may support learning, and research on self-directed learning has been accelerated by integrating education and IT technology. The process of self-directed learning in e-learning applications such as Car Maintenance Training is very difficult and complicated. Previous studies on car maintenance training applications provided simple training scenarios with predetermined action sequences. To incorporate self-directed learning in car maintenance training, however, trainees must be able to perform various maintenance operations himself and experience various situations. To provide such functionality, it is necessary to obtain an accurate response for various operations of trainees, but it requires complicated calculations with respect to varieties in the electrical and mechanical processes of a car. In this paper, we develop a logic simulation agent using JESS inference engine in which self-directed learning is achieved by capturing the behavior of trainees and simulating car operations without complicated physical simulations in car maintenance training.