• Journal of the Korean Data and Information Science Society
• /
• v.24 no.2
• /
• pp.391-399
• /
• 2013

Partially linear regression is capable of providing more complete description of the linear and nonlinear relationships among random variables. In support vector regression (SVR) the hyper-parameters are known to affect the performance of regression. In this paper we propose an iterative reweighted least squares (IRWLS) procedure to solve the quadratic problem of partially linear support vector regression with a modified loss function, which enables us to use the generalized approximate cross validation function to select the hyper-parameters. Experimental results are then presented which illustrate the performance of the partially linear SVR using IRWLS procedure.

• Proceedings of the KIEE Conference
• /
• 2006.04a
• /
• pp.90-92
• /
• 2006

기존의 Support Vector Data Description (SVDD) 방법은 학습 데이터의 개수가 증가함에 따라 학습 시간이 지수 함수적으로 증가하므로, 대량의 데이터를 학습하는 데에는 한계가 있었다. 본 논문에서는 학습 속도를 빠르게 하기 위해 K-means clustering 알고리즘을 이용하는 SVDD 알고리즘을 제안하고자 한다. 제안된 알고리즘은 기존의 decomposition 방법과 유사하게 K-means clustering 알고리즘을 이용하여 학습 데이터 영역을 sub-grouping한 후 각각의 sub-group들을 개별적으로 학습함으로써 계산량 감소 효과를 얻는다. 이러한 sub-grouping 과정은 hypersphere를 이용하여 학습 데이터를 둘러싸는 SVDD의 학습 특성을 훼손시키지 않으면서 중심점으로 모여진 작은 영역의 학습 데이터를 학습하도록 함으로써, 기존의 SVDD와 비교하여 학습 정확도의 차이 없이 빠른 학습을 가능하게 한다. 다양한 데이터들을 이용한 모의실험을 통하여 그 효과를 검증하도록 한다.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.9 no.2
• /
• pp.141-146
• /
• 2009

Support vector machine is a classification algorithm based on statistical learning theory. It has shown many results with good performances in the data mining fields. But there are some problems in the algorithm. One of the problems is its heavy computing cost. So we have been difficult to use the support vector machine in the dynamic and online systems. To overcome this problem we propose to use stratified sampling of statistical sampling theory. The usage of stratified sampling supports to reduce the size of training data. In our paper, though the size of data is small, the performance accuracy is maintained. We verify our improved performance by experimental results using data sets from UCI machine learning repository.

• Phonetics and Speech Sciences
• /
• v.9 no.2
• /
• pp.85-94
• /
• 2017

Conventional speaker verification systems using time delay neural network, identity vector and probabilistic linear discriminant analysis (TDNN-Ivector-PLDA) are known to be very effective for verifying long-duration speech utterances. However, when test utterances are of short duration, duration mismatch between enrollment and test utterances significantly degrades the performance of TDNN-Ivector-PLDA systems. To compensate for the I-vector mismatch between long and short utterances, this paper proposes to use probabilistic linear discriminant analysis (PLDA) model adaptation with augmented data. A PLDA model is trained on vast amount of speech data, most of which have long duration. Then, the PLDA model is adapted with the I-vectors obtained from short-utterance data which are augmented by using vocal tract length perturbation (VTLP). In computer experiments using the NIST SRE 2008 database, the proposed method is shown to achieve significantly better performance than the conventional TDNN-Ivector-PLDA systems when there exists duration mismatch between enrollment and test utterances.

• Journal of the Korean Data and Information Science Society
• /
• v.23 no.2
• /
• pp.393-401
• /
• 2012

This paper deals with the estimations of the least squares support vector regression when the responses are subject to randomly right censoring. The estimation is performed via two steps - the ordinary least squares support vector regression and the least squares support vector regression with censored data. We use the empirical fact that the estimated regression functions subject to randomly right censoring are close to the true regression functions than the observed failure times subject to randomly right censoring. The hyper-parameters of model which affect the performance of the proposed procedure are selected by a generalized cross validation function. Experimental results are then presented which indicate the performance of the proposed procedure.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.23 no.5
• /
• pp.732-736
• /
• 1986

This paper propose a novel vector-quantized block truncation coder for image data compression. A data compression ratio of about 3-6 times larger than that of the BTC can be achieved by utilizign a vector quantizer with the BTC. A vector quantizer was realized by computer simulation. The compressed data rate of 0.7~1.0 bit/pel with intraframe coder and that of 0.3~0.5 bit/pel with interframe coder gives a good performance.

• Journal of Multimedia Information System
• /
• v.4 no.4
• /
• pp.249-254
• /
• 2017

Digital vector maps must be compressed effectively for transmission or storage in Web GIS (geographic information system) and mobile GIS applications. This paper presents a polyline compression method that consists of polyline feature-based hybrid simplification and second derivative-based data compression. Experimental results verify that our method has higher simplification and compression efficiency than conventional methods and produces good quality compressed maps.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.4
• /
• pp.27-36
• /
• 2010

The paper proposes RFB(Record Filtering using Bit-vector) join algorithm, an in-network strategy that uses bit-vector to drastically reduce the size of data and hence the communication cost. In addition, by eliminating data not involved in join result prior to actual join, communication cost can be minimized since not all data need to be moved to the join nodes. The simulation result shows that the proposed RFB algorithm significantly reduces the number of bytes to be moved to join nodes compared to the popular synopsis join(SNJ) algorithm.

• Journal of Korea Multimedia Society
• /
• v.21 no.7
• /
• pp.755-763
• /
• 2018

In this paper, for recognize a night pedestrian from an infrared video, a new method differentiated from the existing feature vector is proposed and experimented. The new approach focuses on the shape feature vector of the structure and shape of the pedestrian image divided by the human body seven split ratio. The pedestrian images are divided into 7 square blocks from the still image of the preprocessing process. And to reduce the dimension, the square block is converted into a mosaic block. The scalar and direction of the shape feature vector is calculated by the brightness and position of the element in the mosaic. For practicality of infrared video system, the proposed method simplifies the data to be processed by reducing the amount of data in the preprocessing in order to continuously batch process the entire system in real time. Through the experiments, we verified the validity of the proposed shape feature vector. In comparison to the existing method, we propose a new shape feature vector generation method as the feature vector for night pedestrian recognition.

• Journal of KIISE:Databases
• /
• v.30 no.3
• /
• pp.273-285
• /
• 2003

Generally, vector spatial data, with richer information than raster spatial data enabledata, enables a mere flexible and effective manipulation of the data sets. However, one of challenges against the publication of vector spatial information on the Internet is the efficient transmission of the big and complex vector spatial datadata, which is both large and complex, across the narrow-bandwidth of the Internet. This paper proposes a new transmission method, namely, the Scale-Dependent Transmission method, with the purpose of improving the efficiency of vector spatial data transmission on the narrow-bandwidthacross the Internet. Simply put, its nam idea is “Transmit what can be seen””. Scale is regarded as a factor naturally associated with spatial features so that not all features are visible to users at a certain scale. With the aid of the Wavelet-Wavelet-based Map Generalization Algorithm, the proposed method filters out invisible features from spatial objects according to the display scale and then to transmit onlytransmits only the visible features as athe final answer for an individual operation. Experiments show that the response times ofan individual operation has been reducedoperations were substantially by the usage of reduced when using the proposed method.