• 대한화학회지
• /
• 제60권4호
• /
• pp.225-234
• /
• 2016

The activity of 34 sulfonamide derivatives has been estimated by means of multiple linear regression (MLR), artificial neural network (ANN), simulated annealing (SA) and genetic algorithm (GA) techniques. These models were also utilized to select the most efficient subsets of descriptors in a cross-validation procedure for non-linear -log (IC50) prediction. The results obtained using GA-ANN were compared with MLR-MLR, MLR-ANN, SA-ANN and GA-ANN approaches. A high predictive ability was observed for the MLR-MLR, MLR-ANN, SA-ANN and MLR-GA models, with root mean sum square errors (RMSE) of 0.3958, 0.1006, 0.0359, 0.0326 and 0.0282 in gas phase and 0.2871, 0.0475, 0.0268, 0.0376 and 0.0097 in solvent, respectively (N=34). The results obtained using the GA-ANN method indicated that the activity of derivatives of sulfonamides depends on different parameters including DP03, BID, AAC, RDF035v, JGI9, TIE, R7e+, BELM6 descriptors in gas phase and Mor 32u, ESpm03d, RDF070v, ATS8m, MATS2e and R4p, L1u and R3m in solvent. In conclusion, the comparison of the quality of the ANN with different MLR models showed that ANN has a better predictive ability.

• Journal of applied mathematics & informatics
• /
• 제34권3_4호
• /
• pp.319-330
• /
• 2016

The hyper-Zagreb index HM(G) of a simple graph G is defined as the sum of the terms (d_u+d_v)² over all edges uv of G, where d_u denotes the degree of the vertex u of G. In this paper, we present several upper and lower bounds on the hyper-Zagreb index in terms of some molecular structural parameters and relate this index to various well-known molecular descriptors.

• Molecular & Cellular Toxicology
• /
• 제5권1호
• /
• pp.14-22
• /
• 2009

In vitro cytotoxicity of 23 alkyl phenols (APs) on human cervical cancer cell lines (HeLa) was determined using the lactate dehydrogenase (LDH) cytotoxicity assay. Two different sets of descriptors were used to construct the calibration model based on Genetic Algorithm-Multiple Linear Regression (GA-MLR) based on the experimental data. A statistically robust Structure-Activity Relationships (QSAR) model was achieved ($R^2$=95.05%, $Q^2_{LOO}$=91.23%, F=72.02 and SE= 0.046) using three Dragon descriptors based on Me (0D-Constitutional descriptor), BELp8 (2D-Burden eigenvalue descriptor) and HATS8p (3D-GETAWAY descriptor). However, external validation could not fully prove its validity of the selected QSAR in characterization of the cytotoxicity of APs towards HeLa cells. Nevertheless, the cytotoxicity profiles showed a finding that 4-n-octylphenol (4-NOP), 4-tert-octyl-phenol (4-TOP), 4-n-nonylphenol (4-NNP) had a more potent cytotoxic effect than other APs tested, inferring that increased length and molecular bulkiness of the substituent had important influence on the LDH cytotoxicity.

• 한국염색가공학회지
• /
• 제21권5호
• /
• pp.21-26
• /
• 2009

In order to analyze numerically the hydrophobicity of the new dyes synthesized for unmodified pure polypropylene fibers, the octanol-water partition coefficient (logP), which is one of molecular descriptors representing hydrophobicity of organic compounds, was obtained by a semi-empirical method using Chem3D software. For the dyes of higher logP than around 5, the affinity of the dyes towards unmodified polypropylene fiber was substantial. In contrast to the new dyes for polypropylene, conventional disperse dyes have logP values lower than 5 and exhibited poor affinity.

• Journal of Information Processing Systems
• /
• 제11권4호
• /
• pp.643-654
• /
• 2015

Face recognition under controlled settings, such as limited viewpoint and illumination change, can achieve good performance nowadays. However, real world application for face recognition is still challenging. In this paper, we propose using the combination of Affine Scale Invariant Feature Transform (SIFT) and Probabilistic Similarity for face recognition under a large viewpoint change. Affine SIFT is an extension of SIFT algorithm to detect affine invariant local descriptors. Affine SIFT generates a series of different viewpoints using affine transformation. In this way, it allows for a viewpoint difference between the gallery face and probe face. However, the human face is not planar as it contains significant 3D depth. Affine SIFT does not work well for significant change in pose. To complement this, we combined it with probabilistic similarity, which gets the log likelihood between the probe and gallery face based on sum of squared difference (SSD) distribution in an offline learning process. Our experiment results show that our framework achieves impressive better recognition accuracy than other algorithms compared on the FERET database.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권2호
• /
• pp.763-774
• /
• 2015

Hand posture recognition has been a wide region of applications in Human Computer Interaction and Computer Vision for many years. The problem arises mainly due to the high dexterity of hand and self-occlusions created in the limited view of the camera or illumination variations. To remedy these problems, a hand posture recognition method using 3-D point cloud is proposed to explicitly utilize 3-D information from depth maps in this paper. Firstly, hand region is segmented by a set of depth threshold. Next, hand image normalization will be performed to ensure that the extracted feature descriptors are scale and rotation invariant. By robustly coding and pooling 3-D facets, the proposed descriptor can effectively represent the various hand postures. After that, SVM with Gaussian kernel function is used to address the issue of posture recognition. Experimental results based on posture dataset captured by Kinect sensor (from 1 to 10) demonstrate the effectiveness of the proposed approach and the average recognition rate of our method is over 96%.

• 전자공학회논문지CI
• /
• 제42권1호
• /
• pp.51-58
• /
• 2005

레이저 스트라이프 신호를 물체에 투사하고 그 반사파형을 비전 카메라를 통하여 입력하면 물체의 입체적 정보를 갖는 파형이 나타나게 된다. 이를 활용하여 물체의 3차인 입체정보를 저장하고 활용하고 있는 상황이다. 본 연구는 사람에 대해 적은 수의 스트라이프 신호를 투사하고 반사된 입체적 스트라이프 파형을 분석하였으며 사람에 대한 특징적 패턴 파형을 추출하고 푸리에 서술자를 이용하여 이러한 패턴을 분류 저장하였다. Baum-Welch 알고리즘을 통하여 저장된 특징 파형을 학습시키고 HMM을 이용하여 사람의 형상에 대한 스트라이프 신호를 컴퓨터에 입력하고 사람의 형상을 인식할 수 있는 지를 실험하였으며 푸리에 서술자를 통한 인식방법과 비교하였다. 실험을 통해서 특정 파형으로 학습 후 인체의 형상을 인식할 수 있음을 확인하였으며, HMM을 통한 인식방법이 우수함을 알 수 있었다.

• Korean Chemical Engineering Research
• /
• 제61권4호
• /
• pp.523-541
• /
• 2023

3D 프린터의 활용이 높아짐에 따라 발생하는 화학물질에 대한 노출 빈도가 증가하고 있다. 그러나 3D 프린팅 발생 화학물질의 독성 및 유해성에 대한 연구는 미비하며, 분자 구조 데이터의 결측치로 인해 in silico 기법을 사용한 독성예측 연구는 저조한 실정이다. 본 연구에서는 화학물질의 분자구조 정보를 나타내는 주요 분자표현자의 결측치를 보간하여 3D 프린팅의 독성 및 유해성을 예측한 Data-centric QSAR 모델을 개발하였다. 먼저 MissForest 알고리즘을 사용해 3D 프린팅으로 발생되는 유해물질의 분자표현자 결측치를 보완하였으며, 서로 다른 4가지 기계학습 모델(결정트리, 랜덤포레스트, XGBoost, SVM)을 기반으로 Data-centric QSAR 모델을 개발하여 생물 농축 계수(Log BCF)와 옥탄올-공기분배계수(Log Koa), 분배계수(Log P)를 예측하였다. 또한, 설명 가능한 인공지능(XAI) 방법론 중 TreeSHAP (SHapley Additive exPlanations) 기법을 활용하여 Data-centric QSAR 모델의 신뢰성을 입증하였다. MissForest 알고리즘 기반 결측지 보간 기법은, 기존 분자구조 데이터에 비하여 약 2.5배 많은 분자구조 데이터를 확보할 수 있었다. 이를 바탕으로 개발된 Data-centric QSAR 모델의 성능은 Log BCF, Log Koa와 Log P를 각각 73%, 76%, 92% 의 예측 성능으로 예측할 수 있었다. 마지막으로 Tree-SHAP 분석결과 개발된 Data-centric QSAR 모델은 각 독성치와 물리적으로 상관성이 높은 분자표현자를 통하여 선택함을 설명할 수 있었고 독성 정보에 대한 높은 예측 성능을 확보할 수 있었다. 본 연구에서 개발한 방법론은 다른 프린팅 소재나 화학공정, 그리고 반도체/디스플레이 공정에서 발생 가능한 오염물질의 독성 및 인체 위해성 평가에 활용될 수 있을 것으로 사료된다.

• 한국정보과학회논문지:시스템및이론
• /
• 제32권11_12호
• /
• pp.705-716
• /
• 2005

패턴 인식 연구 분야에서 많은 특징점 추출 알고리즘들이 개발되었지만, 태블릿 PC나 LCD 태블릿과 같은 펜-입력 디스플레이를 위한 인터액티브 애플리케이션들은 기존과는 다른 요구사항을 가진다. 사용자 마다 다른 다양한 스케치 스타일의 대해서 세그멘테이션 및 특징점 추출을 그림을 그리는 동안 실시간에 안정적으로 수행하여야 한다. 본 논문은 사용자로부터 자유로이 입력된 펜 입력을 분할(segmentation)하기 위해 필수적인 곡률(curvature) 측정 방법을 제안한다. 이 방법은 국소적인 모양 정보(shape descriptors)만을 사용하므로 펜 입력동안 곧바로(on-the-fly) 곡률을 측정할 수 있다. 본 알고리즘은 3차원 스케치 기반 모델링 애플리케이션에서 펜 마킹 인식을 위해서 사용되었다.

• Bulletin of the Korean Chemical Society
• /
• 제30권1호
• /
• pp.67-76
• /
• 2009

The development of a variety of methods like AM1, PM3, PM5 and DFT now allows the calculation of atomic and molecular properties with high precision as well as the treatment of large molecules with predictive power. In this paper, these methods have been used to calculate a number of quantum chemical descriptors (like Klopman atomic softness in terms of $E_n^{\ddag}\;and\;E_m^{\ddag}$, chemical hardness, global softness, electronegativity, chemical potential, electrophilicity index, heat of formation, total energy etc.) for 75 alkanes to predict their boiling point values. The 3D modeling, geometry optimization and semiempirical & DFT calculations of all the alkanes have been made with the help of CAChe software. The calculated quantum chemical descriptors have been correlated with observed boiling point by using multiple linear regression (MLR) analysis. The predicted values of boiling point are very close to the observed values. The values of correlation coefficient ($r^2$) and cross validation coefficient ($r_{cv}^2$) also indicates the generated QSPR models are valuable and the comparison of all the methods indicate that the DFT method is most reliable while the addition of Klopman atomic softness $E_n^{\ddag}$ in DFT method improves the result and provides best correlation.