• Journal of the Korea Society of Computer and Information
• /
• v.17 no.11
• /
• pp.11-18
• /
• 2012

Debris flow deposition model is a model to predict affected areas by debris flow and random walk model (RWM) was used to build the model. Although the model was proved to be effective in the prediction of affected areas, the model has several free parameters decided experimentally. There are several well-known methods to estimate parameters, however, they cannot be applied directly to the debris flow problem due to the small size of training data. In this paper, a modified neural network, called pseudo sample neural network (PSNN), was proposed to overcome the sample size problem. In the training phase, PSNN uses pseudo samples, which are generated using the existing samples. The pseudo samples smooth the solution space and reduce the probability of falling into a local optimum. As a result, PSNN can estimate parameter more robustly than traditional neural networks do. All of these can be proved through the experiments using artificial and real data sets.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.4
• /
• pp.19-26
• /
• 2013

Pseudo sample neural network (PSNN) is a variant of traditional neural network using pseudo samples to mitigate the local-optima-convergence problem when the size of training samples is small. PSNN can take advantage of the smoothed solution space through the use of pseudo samples. PSNN has a focus on the quantity problem in training, whereas, methods stressing the quality of training samples is presented in this paper to improve further the performance of PSNN. It is evident that typical samples and highly correlated features help in training. In this paper, therefore, kernel density estimation is used to select typical samples and correlation factor is introduced to select features, which can improve the performance of PSNN. Debris flow data set is used to demonstrate the usefulness of the proposed methods.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2013.01a
• /
• pp.197-199
• /
• 2013

신경망의 학습은 학습 샘플의 품질뿐만이 아니라 입력으로 사용되는 특징에도 영향을 받으므로 신경망의 출력을 결정하는데 있어 연관성이 높은 특징을 입력으로 사용함으로써 학습된 신경망의 전체적인 성능을 높일 수 있다. 이 논문에서는 신경망의 입력으로 사용되는 특징과 출력의 연관성 파악하고 연관성이 낮은 특징을 학습 과정에서 배제함으로써 신경망의 전체적인 성능을 높일 수 있는 방법을 제시하였다. 토석류 데이터를 위한 의사 샘플 신경망에 제안한 방법을 적용한 경우 연관성이 낮은 특징 하나를 제외함으로써 약 6%의 오류 감소 효과를 얻을 수 있었다.

• Smart Structures and Systems
• /
• v.26 no.1
• /
• pp.63-75
• /
• 2020

Substructure pseudo-dynamic hybrid simulation (SPDHS) combining the advantages of physical experiments and numerical simulation has become an important testing method for evaluating the dynamic responses of structures. Various parameter identification methods have been proposed for online model updating. However, if there is large model gap between the assumed numerical models and the real models, the parameter identification methods will cause large prediction errors. This study presents an ANN (artificial neural network) method based on forgetting factor. During the SPDHS of model updating, a dynamic sample window is formed in each loading step with forgetting factor to keep balance between the new samples and historical ones. The effectiveness and anti-noise ability of this method are evaluated by numerical analysis of a six-story frame structure with BRBs (Buckling Restrained Brace). One BRB is simulated in OpenFresco as the experimental substructure, while the rest is modeled in MATLAB. The results show that ANN is able to present more hysteresis behaviors that do not exist in the initial assumed numerical models. It is demonstrated that the proposed method has good adaptability and prediction accuracy of restoring force even under different loading histories.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2012.07a
• /
• pp.33-34
• /
• 2012

토석류 퇴적 모델은 토석류에 의한 피해지 예측을 위해 그 효용성이 입증된 모델이지만 이를 이용하기 위해서는 몇 가지 파라미터를 필요로 한다. 파라미터를 자동으로 추정하기 위한 방법은 여러 가지가 있지만 토석류에 의한 피해지 예측을 위한 데이터는 충분히 양을 확보하기가 어려우므로 기존의 학습 기법을 적용하는데 어려움이 있다. 본 논문에서는 인공 신경망을 학습시키는 과정에서 기존 샘플로부터 의사 샘플을 생성하고 이를 학습에 사용함으로써 보다 안정적인 학습이 가능한 의사 샘플 신경망을 제안하였다. 제안한 의사 샘플 신경망은 해공간을 평탄화시킴으로써 잘못된 국부 최적해에 빠질 확률을 줄여주고 따라서 보다 안정적인 파라미터 추정이 가능하다는 사실을 실험을 통해 확인할 수 있다.