• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.1
• /
• pp.85-92
• /
• 2020

Deep learning shows outstanding performance in image and video analysis, such as object classification, object detection and semantic segmentation. In this paper, it is analyzed that the performances of deep learning models can be affected by characteristics of train dataset. It is proposed as a method for selecting activation function and optimization algorithm of deep learning to classify facial expression. Classification performances are compared and analyzed by applying various algorithms of each component of deep learning model for CK+, MMI, and KDEF datasets. As results of simulation, it is shown that genetic algorithm can be an effective solution for optimizing components of deep learning model.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.12
• /
• pp.79-87
• /
• 2023

In this paper, we propose a novel genetic algorithm to reduce the overall span time by optimizing the skid insertion sequence in the shipyard subassembly process. We represented a solution by a permutation of a set of skid ids and applied genetic operators suitable for such a representation. In addition, we combined the genetic algorithm and the existing heuristic algorithm called UniDev which is properly modified to improve the search performance. In particular, the slow skid search part in UniDev was changed to a greedy algorithm. Through extensive large-scaled simulations, it was observed that the span time of our method was stably minimized compared to Multi-Start search and a genetic algorithm combined with UniDev.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.37 no.12
• /
• pp.1513-1519
• /
• 2013

A genetic algorithm (GA) is an optimization technique based on natural evolution theory to find the global optimal solution. Unlike the gradient-based method, it can design nanoscale structures in the electric field because it does not require sensitivity calculation. This research intends to design a nanoaperture with an unprecedented shape by the topology optimization scheme based on the GA and ON/OFF method in the visible frequency range. This research mainly aims to maximize the transmission rate at a measuring area located 10nm under the exit plane and to minimize the electric distribution at other locations. The finite element analysis (FEA) and optimization process are performed by using the commercial package COMSOL combined with the Matlab programming. The final results of the optimized model are analyzed by a comparison of the electric field intensity and the spot size of near field with those of the initial model.

• Journal of the Korea Convergence Society
• /
• v.12 no.1
• /
• pp.49-55
• /
• 2021

Recently, with the development of database, it is possible to store a lot of data generated in finance, security, and networks. These data are being analyzed through classifiers based on machine learning. The main problem at this time is data imbalance. When we train imbalanced data, it may happen that classification accuracy is degraded due to over-fitting with majority class data. To overcome the problem of data imbalance, oversampling strategy that increases the quantity of data of minority class data is widely used. It requires to tuning process about suitable method and parameters for data distribution. To improve the process, In this study, we propose a strategy to explore and optimize oversampling combinations and ratio based on various methods such as synthetic minority oversampling technique and generative adversarial networks through genetic algorithms. After sampling credit card fraud detection which is a representative case of data imbalance, with the proposed strategy and single oversampling strategies, we compare the performance of trained classifiers with each data. As a result, a strategy that is optimized by exploring for ratio of each method with genetic algorithms was superior to previous strategies.

• Journal of the Institute of Convergence Signal Processing
• /
• v.14 no.1
• /
• pp.57-61
• /
• 2013

A hybrid genetic algorithm based learning method for the morphological neural networks (MNN) is proposed. The morphological neural networks are based on max-plus algebra, therefore, it is difficult to optimize the coefficients of MNN by the learning method with derivative operations. In order to solve the difficulty, a hybrid genetic algorithm based learning method to optimize the coefficients of MNN is used. Through the image compression/reconstruction experiment using test images extracted from standard image database(SIDBA), it is confirmed that the quality of the reconstructed images obtained by the proposed method is better than that obtained by the conventional neural networks.

• Proceedings of the Korean Information Science Society Conference
• /
• 2005.11a
• /
• pp.721-723
• /
• 2005

본 논문에서는 영상 기반 항해를 위하여 최적의 영상 샘플링을 획득하기 위한 영상 샘플링 알고리즘을 제안한다. 이를 위해 초기 과다 샘플링 된 영상열의 입력으로부터 장면 전역에 걸쳐 적절한 렌더링 품질을 보장하는 최소의 샘플링을 선택하는 감소 샘플링(decremental sampling)의 접근 방법을 기반으로 본 문제를 Set Covering 문제로 공식화한다. 각 시점으로부터 최상의 영상 품질을 보장하는 영역을 3D 와핑 알고리즘을 사용하여 포함 영역으로 표현하여, 이렇게 표현된 Set Covering 문제는 유전 알고리즘을 사용하여 최적화 문제로 설계한다. 실험 결과 본 논문에서 제안한 방법을 통해 최적 해를 구함으로서 만족할 만한 영상 기반 항해의 결과를 얻을 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.10a
• /
• pp.641-644
• /
• 2018

열간 자유 단조는 고온으로 가열한 강피에 압력을 가하여 원하는 형상을 빚는 공정이다. 가열로에서 여러 개의 강피를 동시에 가열하며 목표 온도에 도달하면 꺼내어 다음 공정을 진행한다. 이때 가열로에 투입하는 소재의 조합과 후단 공정을 위해 소재를 꺼내는 순서가 가열로의 에너지 효율에 영향을 끼친다. 본 논문에서는 열간 자유 단조의 에너지 효율을 높이기 위한 비용 예측 모형 기반 작업 계획 최적화 방안을 제안한다. 유전 알고리즘을 이용하여 가열로 강피 조합을 최적화하며 각 설비별 작업 할당 규칙에 따라 전체 작업 계획을 수립한다. 시뮬레이션 기반으로 후보 작업 계획을 평가하여 계획을 최적화 하며 이를 위해 각 설비별 공정 소요 시간 및 에너지 사용량 예측 모형을 이용한다. 예측 모형은 공정 데이터를 기반으로 기계 학습 알고리즘을 적용하여 학습한다. 또한 주기적인 재계획을 통해 예측의 불확실성으로 인해 작업의 진행이 계획대로 이루어지지 않는 문제점을 해결하고자 한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.599-600
• /
• 2023

본 연구는 한국 주식 시장인 KRX 데이터에 중점을 둔 투자 전략의 최적화를 목표로 하였다. 전통적인 기술 분석 방법은 투자자들의 경험에 의존하여 파라미터를 선택하였다. 하지만 이 연구에서는 기존의 경험에 기반한 파라미터 선택 대신 유전 알고리즘을 사용하여 파라미터를 최적화했다. 결과적으로, 이 전략은 상승장과 하락장 모두에서 buy-and-hold 전략보다 더 나은 성과를 보였다. 이는 기술 분석의 파라미터 최적화의 중요성을 강조하며, 더 효과적인 투자 전략 개발의 가능성을 보여준다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10a
• /
• pp.178-180
• /
• 2003

최소 걸침 나무는 널리 알려진 순회 판매원 문제와 같이 전통적인 최적화 문제 중에 하나이다. 특히나 최소 걸침 나무와는 달리 차수 제약 최소 걸침 나무의 경우는 일반적으로 NP-hard 문제로 알려져 있다. 이러한 NP-hard 문제를 해결하기 위한 다양한 접근법들이 소개되었는데 유전 알고리즘은 효율적인 방법 중에 하나로 알려져 있다. 유전 알고리즘과 같이 진화에 기반을 둔 알고리즘을 어떤 문제에 적응하기 위해서 가장 우선적으로 고려되어야 하는 것은 해를 어떻게 표현할 것인가 인데 본 논문에서는 차수 제약 최소 걸침 나무를 해결하기 위한 새로운 트리 표현법을 제안한다.

• Journal of the Korea Convergence Society
• /
• v.10 no.2
• /
• pp.13-20
• /
• 2019

For customized education, it is essential to search the learning path for the learner. The genetic algorithm makes it possible to find optimal solutions within a practical time when they are difficult to be obtained with deterministic approaches because of the problem's very large search space. In this research, based on genetic algorithm, the learning paths to learn 200 chords in 27 music sheets were optimized to maximize the learning effect by balancing and minimizing learner's burden and learning size for each step in the learning paths. Although the permutation size of the possible learning path for 27 learning contents is more than $10^{28}$, the optimal solution could be obtained within 20 minutes in average by an implemented tool in this research. Experimental results showed that genetic algorithm can be effectively used to design complex learning path for customized education with various purposes. The proposed method is expected to be applied in other educational domains as well.