• 대한전기학회논문지:시스템및제어부문D
• /
• 제53권7호
• /
• pp.531-535
• /
• 2004

A new method is presented to construct a classifier. This was accomplished by combining a generalized regression neural network (GRNN) and a genetic algorithm (GA). The classifier constructed in this way is referred to as a GA-GRNN. The GA played a role of controlling training factors simultaneously. The GA-GRNN was applied to classify 4 different Promoter sequences. The training and test data were composed of 115 and 58 sequence patterns, respectively. The classifier performance was investigated in terms of the classification sensitivity and prediction accuracy. Compared to conventional GRNN, GA-GRNN significantly improved the total classification sensitivity as well as the total prediction accuracy. As a result, the proposed GA-GRNN demonstrated improved classification sensitivity and prediction accuracy over the convention GRNN.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1996년도 추계학술대회 학술발표 논문집
• /
• pp.27-34
• /
• 1996

GENIE is a learning-based engine for building intelligent systems. Learning in GENIE proceeds by incrementally modeling its human or technical environment using a neural network and a genetic algorithm. The neural network is used to represent the knowledge for solving a given task and has the ability to grow its structure. The genetic algorithm provides the neural network with training examples by actively exploring the example space of the problem. Integrated into the training examples by actively exploring the example space of the problem. Integrated into the GENIE system architecture, the genetic algorithm and the neural network build a virtually self-teaching autonomous learning system. This paper describes the structure of GENIE and its learning components. The performance is demonstrated on a robot learning problem. We also discuss the lessons learned from experiments with GENIE and point out further possibilities of effectively hybridizing genetic algorithms with neural networks and other softcomputing techniques.

• Journal of Computational Design and Engineering
• /
• 제2권3호
• /
• pp.137-147
• /
• 2015

In this study, a new methodology in predicting a system output has been investigated by applying a data mining technique and a hybrid type II fuzzy system in CNC turning operations. The purpose was to generate a supplemental control function under the dynamic machining environment, where unforeseeable changes may occur frequently. Two different types of membership functions were developed for the fuzzy logic systems and also by combining the two types, a hybrid system was generated. Genetic algorithm was used for fuzzy adaptation in the control system. Fuzzy rules are automatically modified in the process of genetic algorithm training. The computational results showed that the hybrid system with a genetic adaptation generated a far better accuracy. The hybrid fuzzy system with genetic algorithm training demonstrated more effective prediction capability and a strong potential for the implementation into existing control functions.

• 대한전기학회논문지:전력기술부문A
• /
• 제54권6호
• /
• pp.259-266
• /
• 2005

Short-term load forecasting algorithm using neural networks and the sensitivity of temperatures in the summer season is proposed. In recent 10 years, many researchers have focused on artificial neural network approach for the load forecasting. In order to improve the accuracy of the load forecasting, input parameters of neural networks are investigated for three training cases of previous 7-days, 14-days, and 30-days. As the result of the investigation, the training case of previous 7-days is selected in the proposed algorithm. Test results show that the proposed algorithm improves the accuracy of the load forecasting.

• Structural Engineering and Mechanics
• /
• 제49권6호
• /
• pp.687-703
• /
• 2014

This paper develops a new nonlinear model for active control of three-dimensional (3D) irregular building structures. Both geometrical and material nonlinearities with a neuro-controller training algorithm are applied to a multi-degree-of-freedom 3D system. Two dynamic assembling motions are considered simultaneously in the control model such as coupling between torsional and lateral responses of the structure and interaction between the structural system and the actuators. The proposed control system and training algorithm of the structural system are evaluated by simulating the responses of the structure under the El-Centro 1940 earthquake excitation. In the numerical example, the 3D three-story structure with linear and nonlinear stiffness is controlled by a trained neural network. The actuator dynamics, control time delay and incident angle of earthquake are also considered in the simulation. Results show that the proposed control algorithm for 3D buildings is effective in structural control.

• 대한임베디드공학회논문지
• /
• 제17권4호
• /
• pp.209-215
• /
• 2022

This paper proposes a novel approach for training an unsupervised depth estimation algorithm. The objective of unsupervised depth estimation is to estimate pixel-wise distances from camera without external supervision. While most previous works focus on model architectures, loss functions, and masking methods for considering dynamic objects, this paper focuses on the training framework to effectively use depth cue. The main loss function of unsupervised depth estimation algorithms is known as the photometric error. In this paper, we claim that direct depth cue is more effective than the photometric error. To obtain the direct depth cue, we adopt the technique of knowledge distillation which is a teacher-student learning framework. We train a teacher network based on a previous unsupervised method, and its depth predictions are utilized as pseudo labels. The pseudo labels are employed to train a student network. In experiments, our proposed algorithm shows a comparable performance with the state-of-the-art algorithm, and we demonstrate that our teacher-student framework is effective in the problem of unsupervised depth estimation.

• Journal of Information Processing Systems
• /
• 제19권4호
• /
• pp.540-553
• /
• 2023

The hollow letter CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) is an optimized version of solid CAPTCHA, specifically designed to weaken characteristic information and increase the difficulty of machine recognition. Although convolutional neural networks can solve CAPTCHA in a single step, a good attack result heavily relies on sufficient training data. To address this challenge, we propose a seed filling algorithm that converts hollow characters to solid ones after contour line restoration and applies three rounds of detection to remove noise background by eliminating noise blocks. Subsequently, we utilize a support vector machine to construct a feature vector for recognition. Security analysis and experiments show the effectiveness of this algorithm during the pre-processing stage, providing favorable conditions for subsequent recognition tasks and enhancing the accuracy of recognition for hollow CAPTCHA.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권1호
• /
• pp.413-435
• /
• 2018

Although the accuracy of handwritten character recognition based on deep networks has been shown to be superior to that of the traditional method, the use of an overly deep network significantly increases time consumption during parameter training. For this reason, this paper took the training time and recognition accuracy into consideration and proposed a novel handwritten character recognition algorithm with newly designed network structure, which is based on an extended nonlinear kernel residual network. This network is a non-extremely deep network, and its main design is as follows:(1) Design of an unsupervised apriori algorithm for intra-class clustering, making the subsequent network training more pertinent; (2) presentation of an intermediate convolution model with a pre-processed width level of 2;(3) presentation of a composite residual structure that designs a multi-level quick link; and (4) addition of a Dropout layer after the parameter optimization. The algorithm shows superior results on MNIST and SVHN dataset, which are two character benchmark recognition datasets, and achieves better recognition accuracy and higher recognition efficiency than other deep structures with the same number of layers.

• 한국음향학회지
• /
• 제13권1호
• /
• pp.5-13
• /
• 1994

본 논문은 음소인식을 위한 신경회로망에 관한 연구로서, 시간 지연 신경회로망을 이용하여 음소인식을 수행하였다. 또한, 본 논문은 대규모 시간지연 신경망에도 적합한 음성 인식 신경망의 학습 방법에 제안한다. 연속 음성의 인식을 위해 반드시 선행되어야 하는 음소의 정확한 인식을 위하여 우수한 성능을 보이고 있는 시간지연 신경망을 사용하였으며, 인식 대상 음소수가 증가하여도 신경망을 최적으로 수렴시킬 수 있는 시간지연 신경망의 새로운 알고리즘을 제시하였다. 확률론적 접근법인 코우쉬 알고리즘을 에러 역전파 알고리즘에 결합하는 시간지연 신경망의 새로운 학습 알고리즘을 사용한 실험이 수행되었다. 화자 2인을 대상으로 한 3분류의 음소군 인식 실험에서 $98.1\%$의 인식률을 얻었으며, 제안된 알고리즘이 시간지연 신경망의 더욱 우수한 인식률과 수렴 시간의 단축에 효율적이었음을 보였다.

• 한국도로학회논문집
• /
• 제14권2호
• /
• pp.11-17
• /
• 2012

본 연구는 실험계획법(예: 반응표면계획법) 및 하모니 검색 알고리즘을 통하여 다양한 아스팔트 콘크리트 포장 구조체에 있어 피로균열의 공용성 인자인 인장변형률을 예측하는 모델을 개발하는 방법에 대한 연구이다. 인장변형률을 산정하기 위하여 한국건설기술연구소에서 개발한 유한요소 축대칭해석 프로그램인 KICTPAVE를 이용하여 아스팔트 층과 린콘크리트 층의 접속면에서 발생되는 변형률을 구하여 데이터베이스(D/B)화 하였다. 아스팔트 포장에서 입력변수인 층별 탄성계수 및 두께를 다양한 조건에서 KICTPAVE 프로그램을 수행하여 훈련용 D/B(Training Set)인 변형률의 값들을 구축한 후 반응표면계획법에 근거하여 회귀방정식을 정의하였으며 방정식에 필요한 계수값을 결정하기 위하여 하모니 검색 알고리즘을 이용하였다. 최종적으로 결정된 회귀방정식의 계수값들의 정확성을 검증하기 위해서 훈련용 D/B가 아닌 다른 조건의 입력변수를 이용하여 검증용 D/B(Testing Set)를 구축하고 이를 이용하여 개발된 모델을 검증하였다.