http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Kernel-based Censored Varying Coefficient Regression Using Miller’s Method
Changha Hwang 계명대학교 자연과학연구소 2021 Quantitative Bio-Science Vol.40 No.2
The censored regression model generally assumes the logarithm of survival time is modelled linearly in the covariates. In this study a censored varying coefficient regression model is proposed to consider situations in which the regression coefficients are not constant and change as the relevant smoothing variables change. Using the formulation of weighted least squares support vector machine with jumps of the Kaplan-Meier estimator of the empirical distribution of function of residuals similar to Miller’s estimation for censored regression, we can easily obtain the estimators of the proposed model through simple linear equations, and can also easily derive a generalized cross validation function. The proposed method is evaluated through simulated and real data sets.
Changha Hwang,Jooyong Shim 한국데이터정보과학회 2016 한국데이터정보과학회지 Vol.27 No.3
In this paper, we propose a deep least squares support vector machine (LS-SVM) for regression problems, which consists of the input layer and the hidden layer. In the hidden layer, LS-SVMs are trained with the original input variables and the perturbed responses. For the final output, the main LS-SVM is trained with the outputs from LS-SVMs of the hidden layer as input variables and the original responses. In contrast to the multilayer neural network (MNN), LS-SVMs in the deep LS-SVM are trained to minimize the penalized objective function. Thus, the learning dynamics of the deep LS-SVM are entirely different from MNN in which all weights and biases are trained to minimize one final error function. When compared to MNN approaches, the deep LS-SVM does not make use of any combination weights, but trains all LS-SVMs in the architecture. Experimental results from real datasets illustrate that the deep LS- SVM significantly outperforms state of the art machine learning methods on regression problems.
Oxidation of organic contaminants in water by iron-induced oxygen activation: A short review
Changha Lee 대한환경공학회 2015 Environmental Engineering Research Vol.20 No.3
Reduced forms of iron, such as zero-valent ion (ZVI) and ferrous ion (Fe[II]), can activate dissolved oxygen in water into reactive oxidants capable of oxidative water treatment. The corrosion of ZVI (or the oxidation of (Fe[II]) forms a hydrogen peroxide (H₂O₂) intermediate and the subsequent Fenton reaction generates reactive oxidants such as hydroxyl radical (●OH) and ferryl ion (Fe[IV]). However, the production of reactive oxidants is limited by multiple factors that restrict the electron transfer from iron to oxygen or that lead the reaction of H₂O₂ to undesired pathways. Several efforts have been made to enhance the production of reactive oxidants by iron-induced oxygen activation, such as the use of iron-chelating agents, electron-shuttles, and surface modification on ZVI. This article reviews the chemistry of oxygen activation by ZVI and Fe(II) and its application in oxidative degradation of organic contaminants. Also discussed are the issues which require further investigation to better understand the chemistry and develop practical environmental technologies.
Deep Multimodal Classification Model for Predicting Successes and Failures of Clinical Trials
Changha Hwang 계명대학교 자연과학연구소 2020 Quantitative Bio-Science Vol.39 No.1
From the discovery of new drug candidates through clinical trials to their approval, it takes approximately 15 years to launch a new drug into the market, and costs approximately one trillion to two trillion won. Despite several improvements in the drug development pipeline over the past 30 years, failures have skyrocketed at all stages of clinical trials owing to safety reasons. To improve the success rate of clinical trials, it is necessary to identify drug candidates that may fail in the clinical trials. Therefore, we need to develop reliable models to predict the outcomes of clinical trials of drug candidates. In this paper, we propose a deep multimodal classification model based on informative chemical features of the drugs and target-based features. Experimental results reported on the PrOCTOR dataset indicate that the proposed model performs better in a multimodal setting. Comparing ensemble models based on random forests and extra trees, the proposed deep multimodal classifier obtains the highest value for the area under the receiver operator curve and area under the precision-recall curve. The results of this study demonstrate that the proposed multimodal classifier can be used to predict the outcomes of clinical trials effectively.