• Genomics & Informatics
• /
• 제2권1호
• /
• pp.19-29
• /
• 2004

Gene expression profiles may offer more information than morphology and provide an alternative to morphology- based tumor classification systems. Informative gene selection is finding gene subsets that are able to discriminate between tumor types, and may have clear biological interpretation. Gene selection is a fundamental issue in gene expression based tumor classification. In this report, techniques for selecting informative genes are illustrated and supervised shaving introduced as a gene selection method in the place of a clustering algorithm. The supervised shaving method showed good performance in gene selection and classification, even though it is a clustering algorithm. Almost selected genes are related to leukemia disease. The expression profiles of 3051 genes were analyzed in 27 acute lymphoblastic leukemia and 11 myeloid leukemia samples. Through these examples, the supervised shaving method has been shown to produce biologically significant genes of more than $94\%$ accuracy of classification. In this report, SVM has also been shown to be a practicable method for gene expression-based classification.

• Geomechanics and Engineering
• /
• 제37권3호
• /
• pp.263-277
• /
• 2024

Artificial Neural Networks (ANNs) are artificial learning algorithms that provide successful results in solving many machine learning problems such as classification, prediction, object detection, object segmentation, image and video classification. There is an increasing number of studies that use ANNs as a prediction tool in soil classification. The aim of this research was to understand the role of hyperparameter optimization in enhancing the accuracy of ANNs for soil type classification. The research results has shown that the hyperparameter optimization and hyperparamter optimized ANNs can be utilized as an efficient mechanism for increasing the estimation accuracy for this problem. It is observed that the developed hyperparameter tool (HyperNetExplorer) that is utilizing the Covariance Matrix Adaptation Evolution Strategy (CMAES), Genetic Algorithm (GA) and Jaya Algorithm (JA) optimization techniques can be successfully used for the discovery of hyperparameter optimized ANNs, which can accomplish soil classification with 100% accuracy.

• International Journal of Automotive Technology
• /
• 제7권7호
• /
• pp.827-832
• /
• 2006

Occupant classification and position detection have been significant research areas in intelligent safety systems in the automotive field. The detection and classification of seat occupancy open up new ways to control the safety system. This paper deals with a novel algorithm development, hardware implementation and testing of a prototype intelligent safety system for occupant classification and position detection for in-vehicle environment. Borland C++ program is used to develop the novel algorithm interface between the sensor and data acquisition system. MEMS strain gauge hermatic pressure sensor containing micromachined integrated circuits is installed inside the passenger seat. The analog output of the sensor is connected with a connector to a PCI-9111 DG data acquisition card for occupancy detection, classification and position detection. The algorithm greatly improves the detection of whether an occupant is present or absent, and the classification of either adult, child or non-human object is determined from weights using the sensor. A simple computation algorithm provides the determination of the occupant's appropriate position using centroidal calculation. A real time operation is achieved with the system. The experimental results demonstrate that the performance of the implemented prototype is robust for occupant classification and position detection. This research may be applied in intelligent airbag design for efficient deployment.

• 한국멀티미디어학회논문지
• /
• 제23권9호
• /
• pp.1129-1138
• /
• 2020

The performance of convolutional deep learning networks is generally determined according to parameters of target dataset, structure of network, convolution kernel, activation function, and optimization algorithm. In this paper, a genetic algorithm is used to select the appropriate deep learning model and parameters for Alzheimer's classification and to compare the learning results with preliminary experiment. We compare and analyze the Alzheimer's disease classification performance of VGG-16, GoogLeNet, and ResNet to select an effective network for detecting AD and MCI. The simulation results show that the network structure is ResNet, the activation function is ReLU, the optimization algorithm is Adam, and the convolution kernel has a 3-dilated convolution filter for the accuracy of dementia medical images.

• 대한의용생체공학회:의공학회지
• /
• 제28권5호
• /
• pp.665-675
• /
• 2007

The morphological change of ECG is the important diagnostic parameter to finding the malfunction of a heart. Generally ST segment deviation is concerned with myocardial abnormality. The aim of this study is to detect the change of ST in shape using a polynomial approximation method and the reference ST type. The developed algorithm consists of feature point detection, ST level detection and ST shape classification. The detection of QRS complex is accomplished using it's the morphological characteristics such as the steep slope and high amplitude. The developed algorithm detects the ST level change, and then classifies the ST shape type using the polynomial approximation. The algorithm finds the least squares curve for the data between S wave and T wave in ECG. This curve is used for the classification of the ST shapes. ST type is classified by comparing the slopes of the specified points between the reference ST set and the least square curve. Through the result from the developed algorithm, we can know when the ST level change occurs and what the ST shape type is.

• Korean Journal of Geomatics
• /
• 제5권1호
• /
• pp.27-34
• /
• 2005

Unsupervised classification is an important area of research in image processing because supervised classification has the disadvantages such as long task-training time and high cost and low objectivity in training information. This paper focuses on unsupervised classification, which can extract ground object information with the minimum 'Spectral Angle Distance' operation on be behalf of 'Spectral Euclidian Distance' in the clustering process. Unlike previous studies, our algorithm uses the unit vector, not the spectral distance, to compute the cluster mean, and the Single-Pass algorithm automatically determines the seed points. Atmospheric correction for more accurate results was adapted on the Hyperion data and the results were analyzed. We applied the algorithm to the Hyperion and ETM+ data and compared the results with K-Means and the former USAM algorithm. From the result, USAM classified the water and dark forest area well and gave more accurate results than K-Means, so we believe that the 'Spectral Angle' can be one of the most accurate classifiers of not only multispectral images but hyperspectral images. And also the unit vector can be an efficient technique for characterizing the Remote Sensing data.

• IEIE Transactions on Smart Processing and Computing
• /
• 제4권4호
• /
• pp.202-208
• /
• 2015

In this study, we propose a new inference algorithm for a multiclass Gaussian process classification model using a variational EM framework and the Laplace approximation (LA) technique. This is performed in two steps, called expectation and maximization. First, in the expectation step (E-step), using Bayes' theorem and the LA technique, we derive the approximate posterior distribution of the latent function, indicating the possibility that each observation belongs to a certain class in the Gaussian process classification model. In the maximization step, we compute the maximum likelihood estimators for hyper-parameters of a covariance matrix necessary to define the prior distribution of the latent function by using the posterior distribution derived in the E-step. These steps iteratively repeat until a convergence condition is satisfied. Moreover, we conducted the experiments by using synthetic data and Iris data in order to verify the performance of the proposed algorithm. Experimental results reveal that the proposed algorithm shows good performance on these datasets.

• 자동차안전학회지
• /
• 제13권4호
• /
• pp.7-13
• /
• 2021

This paper presents an vulnerable road user (VRU) classification and tracking algorithm using vision and LiDAR sensor fusion method for urban autonomous driving. The classification and tracking for vulnerable road users such as pedestrian, bicycle, and motorcycle are essential for autonomous driving in complex urban environments. In this paper, a real-time object image detection algorithm called Yolo and object tracking algorithm from LiDAR point cloud are fused in the high level. The proposed algorithm consists of four parts. First, the object bounding boxes on the pixel coordinate, which is obtained from YOLO, are transformed into the local coordinate of subject vehicle using the homography matrix. Second, a LiDAR point cloud is clustered based on Euclidean distance and the clusters are associated using GNN. In addition, the states of clusters including position, heading angle, velocity and acceleration information are estimated using geometric model free approach (GMFA) in real-time. Finally, the each LiDAR track is matched with a vision track using angle information of transformed vision track and assigned a classification id. The proposed fusion algorithm is evaluated via real vehicle test in the urban environment.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2003년도 신호처리소사이어티 추계학술대회 논문집
• /
• pp.405-408
• /
• 2003

This paper proposes a new algorithm for cloudy area detection using K-Means and GHA (Generalized Hebbian Algorithm). K-Means is one of simple classification algorithm, and GHA is unsupervised neural network for data compression and pattern classification. Proposed algorithm is based on block based image processing that size is l6$\times$l6. Experimental results shows good performance of cloudy area detection except blur cloudy areas.

• 한국통신학회논문지
• /
• 제14권4호
• /
• pp.348-359
• /
• 1989

본 논문은 [1]와 [2]에 의해 제안된 multitree 형상 인식 기법의 성능 개선에 관한 논문이다. Multitree 형상 인식 기법의 기본적인 생각은, Classifier 설계과정에서 각 특징별로 Binary Decision Tree 를 구성하고, 이들의 탐색 순서를 결정하며, 인식 과정에서는 앞에서 정한 탐색 순서에 의거하여, BDT(Binary Decision Tree)를 탐색해 나간다는 것이다. 이때 BDT를 추가하여 탐색하기 전에 그때까지 얻은 정보를 이용하여 입력 물체를 인식할 수 있는지에 대한 여부를 결정하며, 인식이 가능한 경우 BDT의 탐색을 멈추고, 인식이 불가능한 경우 BDT의 탐색을 계속해 나간다. 이 방법은 BDT를 각 특징별로 만들기 때문에 새로운 특징의 삭제나 첨가가 상당히 용이하며 인식에 사용되는 특징의 갯수가 감소하게 된다. 따라서 이 알고리즘은 특징의 수가 많거나 class수가 많을 경우 쉽게 이용될 수 있다. 본 논문은 각 특징에서 구한 근사화된 확률 분포로부터 입력 특징값에 대한 확률값을 구해 인식에 이용하였으며, 이 값을 이용한ㄴ 여러가지 인식 방법을 제안하였다. 그리고 Branch and Bound 방법을 사용하여 특징의 선택 순서와 탐색 범위를 구하였다. 위에서 제안한 것들을 실험한 결과 기존의 multitree형상 인식 기법보다 본 논문에서 제안한 기법의 성능이 향상되었다.