• Proceedings of the Korean Information Science Society Conference
• /
• 2012.06b
• /
• pp.336-338
• /
• 2012

본 연구에서는 고차 데이터 분류를 위해 순차적 베이지만 샘플링 기반의 진화연산 기법을 이용한 하이퍼네트워크 모델의 학습 알고리즘을 제시한다. 제시하는 방법에서는 모델의 조건부 확률의 사후(posterior) 분포를 최대화하도록 학습이 진행된다. 이를 위해 사전(prior) 분포를 문제와 관련된 사전지식(prior knowledge) 및 모델 복잡도(model complexity)로 정의하고, 측정된 모델의 분류성능을 우도(likelihood)로 사 용하며, 측정된 사전분포와 우도를 이용하여 모델의 적합도(fitness)를 정의한다. 이를 통해 하이퍼네트워크 모델은 고차원 데이터를 효율적으로 학습 가능할 뿐이 아니라 모델의 학습시간 및 분류성능이 개선될 수 있다. 또한 학습 시에 파라미터로 주어지던 하이퍼에지의 구성 및 모델의 크기가 학습과정 중에 적응적으로 결정될 수 있다. 제안하는 학습방법의 검증을 위해 본 논문에서는 약 25,000개의 유전자 발현정보 데이터셋에 대한 분류문제에 모델을 적용한다. 실험 결과를 통해 제시하는 방법이 기존 하이퍼네트워크 학습 방법 뿐 아니라 다른 모델들에 비해 우수한 분류 성능을 보여주는 것을 확인할 수 있다. 또한 다양한 실험을 통해 사전분포로 사용된 사전지식이 모델 학습에 끼치는 영향을 분석한다.

• The KIPS Transactions:PartB
• /
• v.10B no.6
• /
• pp.579-586
• /
• 2003

In this paper, we propose a novel approach to employ emotional features to document retrieval systems. Fine emotional features, such as HAPPY, SAD, ANGRY, FEAR, and DISGUST, have been used to represent Korean document. Users are allowed to use these features for retrieving their documents. Next, retrieved documents are learned by classification methods like cohesion factor, naive Bayesian, and, k-nearest neighbor approaches. In order to combine various approaches, voting method has been used. In addition, k-means clustering has been used for our experimentation. The performance of our approach proved to be better in accuracy than other methods, and be better in short texts rather than large documents.

• Journal of the Korean Data and Information Science Society
• /
• v.22 no.2
• /
• pp.197-206
• /
• 2011

The ability to detect online abnormal events in signals is essential in many real-world signal processing applications. In order to detect abnormal events, previously known algorithms require an explicit signal statistical model, and interpret abnormal events as statistical model abrupt changes. In general, maximum likelihood and Bayesian estimation theory to estimate well as detection methods have been used. However, the above-mentioned methods for robust and tractable model, it is not easy to estimate. More freedom to estimate how the model is needed. In this paper, we investigate a machine learning, descriptor-based approach that does not require a explicit descriptors statistical model, based on support vector machines are known to be robust statistical models and a sequential optimal algorithm online support vector machine is introduced.

• Proceedings of the KIEE Conference
• /
• 2004.05a
• /
• pp.126-128
• /
• 2004

As e-mail is becoming increasingly important in every day life activity, mail users spend more and more time organizing and classifying the e-mails they receive into folder. Many existing recommendation systems or text classification are mostly focused on recommending the products for the commercial purposes or web documents. So this study aims to apply these application to e-mail more necessary to users. This paper suggests a dynamic recommendation agent system based on Rule Filtering Component recommending the relevant category to enable users directly to manage the optimum classification when a new e-mail is received as the effective method for E-Mail Management. Moreover we try to improve the accuracy as eliminating the limits of misclassification that can be key in classifying e-mails by category. While the existing Bayesian Learning Algorithm mostly uses the fixed threshold, we prove to improve the satisfaction of users as increasing the accuracy by changing the fixed threshold to the dynamic threshold. We designed main modules by rule filtering component for enhanced scalability and reusability of our system.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.340-340
• /
• 2019

지구온난화로 유발되는 기후변동성이 증가함에 따라서 정확한 수문인자의 예측은 전 세계적으로 주요 관심사항이 되고 있다. 최근에는 고성능 컴퓨터 자원의 증가로 수문기상학 연구에서 동일한 학습량에 비하여 정확도의 향상이 뚜렷한 기계학습 구조를 활용하여 위성영상 기반의 대기예측, 태풍위치 추적 및 강수량 예측 등의 연구가 활발하게 진행되고 있다. 본 연구에는 기계학습 중 시계열 분석에 널리 활용되고 있는 순환신경망(Recurrent Neural Network, RNN) 기법의 대표적인 LSTM(Long Short-Term Memory) 네트워크를 이용하여 수문인자를 예측하였다. LSTM 네트워크는 가중치 및 메모리 요소에 대한 추가정보를 셀 상태에 저장하고 시계열의 길이 조정하여 모형의 탄력적 활용이 가능하다. LSTM 네트워크를 이용한 다양한 수문인자 예측결과 RMSE의 개선을 확인하였다. 따라서 본 연구를 통하여 개발된 기계학습을 통한 수문인자 예측기술은 권역별 수계별 홍수 및 가뭄대응 계획을 능동적으로 수립하는데 활용될 것으로 판단된다. 향후 연구에서는 LSTM의 입력영역을 Bayesian 추론기법을 활용하여 구성함으로 학습과정의 불확실성을 정량적으로 제어하고자 한다.

• Geomechanics and Engineering
• /
• v.28 no.4
• /
• pp.359-374
• /
• 2022

Geology conditions are crucial in decision-making during the planning and design phase of a tunnel project. Estimation of the geology conditions of road tunnels is subject to significant uncertainties. In this work, the effectiveness of a novel regression method in estimating geological or geotechnical parameters of road tunnel projects was explored. This method, called Gaussian process regression (GPR), formulates the learning of the regressor within a Bayesian framework. The GPR model was trained with data of old tunnel projects. To verify its feasibility, the GPR technique was applied to a road tunnel to predict the state of three geological/geomechanical parameters of Rock Mass Rating (RMR), Rock Structure Rating (RSR) and Q-value. Finally, in order to validate the GPR approach, the forecasted results were compared to the field-observed results. From this comparison, it was concluded that, the GPR is presented very good predictions. The R-squared values between the predicted results of the GPR vs. field-observed results for the RMR, RSR and Q-value were obtained equal to 0.8581, 0.8148 and 0.8788, respectively.

• Advances in nano research
• /
• v.12 no.5
• /
• pp.489-499
• /
• 2022

Nanofluids have recently triggered a substantial scientific interest as cooling media. However, their stability is challenging for successful engagement in industrial applications. Different factors, including temperature, nanoparticles and base fluids characteristics, pH, ultrasonic power and frequency, agitation time, and surfactant type and concentration, determine the nanofluid stability regime. Indeed, it is often too complicated and even impossible to accurately find the conditions resulting in a stabilized nanofluid. Furthermore, there are no empirical, semi-empirical, and even intelligent scenarios for anticipating the stability of nanofluids. Therefore, this study introduces a straightforward and reliable intelligent classifier for discriminating among the stability regimes of alumina-water nanofluids based on the Zeta potential margins. In this regard, various intelligent classifiers (i.e., deep learning and multilayer perceptron neural network, decision tree, GoogleNet, and multi-output least squares support vector regression) have been designed, and their classification accuracy was compared. This comparison approved that the multilayer perceptron neural network (MLPNN) with the SoftMax activation function trained by the Bayesian regularization algorithm is the best classifier for the considered task. This intelligent classifier accurately detects the stability regimes of more than 90% of 345 different nanofluid samples. The overall classification accuracy and misclassification percent of 90.1% and 9.9% have been achieved by this model. This research is the first try toward anticipting the stability of water-alumin nanofluids from some easily measured independent variables.

• Journal of Information Processing Systems
• /
• v.18 no.2
• /
• pp.188-196
• /
• 2022

Recently, neural architecture search (NAS) has received increasing attention as it can replace human experts in designing the architecture of neural networks for different tasks and has achieved remarkable results in many challenging tasks. In this study, a path-based computation neural architecture encoder (PCE) was proposed. Our PCE first encodes the computation of information on each path in a neural network, and then aggregates the encodings on all paths together through an attention mechanism, simulating the process of information computation along paths in a neural network and encoding the computation on the neural network instead of the structure of the graph, which is more consistent with the computational properties of neural networks. We performed an extensive comparison with eight encoding methods on two commonly used NAS search spaces (NAS-Bench-101 and NAS-Bench-201), which included a comparison of the predictive capabilities of performance predictors and search capabilities based on two search strategies (reinforcement learning-based and Bayesian optimization-based) when equipped with different encoders. Experimental evaluation shows that PCE is an efficient encoding method that effectively ranks and predicts neural architecture performance, thereby improving the search efficiency of neural architectures.

• International Journal of Computer Science & Network Security
• /
• v.23 no.1
• /
• pp.53-63
• /
• 2023

Many researchers are trying hard to minimize the incidence of cancers, mainly Gastric Cancer (GC). For GC, the five-year survival rate is generally 5-25%, but for Early Gastric Cancer (EGC), it is almost 90%. Predicting the onset of stomach cancer based on risk factors will allow for an early diagnosis and more effective treatment. Although there are several models for predicting stomach cancer, most of these models are based on unbalanced datasets, which favours the majority class. However, it is imperative to correctly identify cancer patients who are in the minority class. This research aims to apply three class-balancing approaches to the NHS dataset before developing supervised learning strategies: Oversampling (Synthetic Minority Oversampling Technique or SMOTE), Undersampling (SpreadSubsample), and Hybrid System (SMOTE + SpreadSubsample). This study uses Naive Bayes, Bayesian Network, Random Forest, and Decision Tree (C4.5) methods. We measured these classifiers' efficacy using their Receiver Operating Characteristics (ROC) curves, sensitivity, and specificity. The validation data was used to test several ways of balancing the classifiers. The final prediction model was built on the one that did the best overall.

• Geomechanics and Engineering
• /
• v.33 no.6
• /
• pp.611-624
• /
• 2023

Soil properties make it attractive as a building material due to its mechanical strength, aesthetically appearance, plasticity, and low cost. However, it is frequently necessary to improve and stabilize the soil mechanical properties with binders. Soil-cement is applied for purposes ranging from housing to dams, roads and foundations. Unconfined compression strength (UCS) and split tensile strength (CD) are essential mechanical parameters for ascertaining the aptitude of soil-cement for a given application. However, quantifying these parameters requires specimen preparation, testing, and several weeks. Methodologies that allowed accurate estimation of mechanical parameters in shorter time would represent an important advance in order to ensure shorter deliverable timeline and reduce the amount of laboratory work. In this work, an extensive campaign of UCS and CD tests was carried out in a sandy soil from the Leiria region (Portugal). Then, using the machine learning tool Neural Pattern Recognition of the MATLAB software, a prediction of these two parameters based on six input parameters was made. The results, especially those obtained with resource to a Bayesian regularization-backpropagation algorithm, are frankly positive, with a forecast success percentage over 90% and very low root mean square error (RMSE).