• Management Science and Financial Engineering
• /
• v.16 no.1
• /
• pp.81-117
• /
• 2010

Semi-supervised learning incorporates unlabeled examples, whose labels are unknown, as well as labeled examples into learning process. Although transductive support vector machine (TSVM), one of semi-supervised learning models, was proposed about a decade ago, its application to large-scaled data has still been limited due to its high computational complexity. Our previous research addressed this limitation by introducing a branch-and-bound algorithm for finding an optimal solution to TSVM. In this paper, we propose three new techniques to enhance the performance of the branch-and-bound algorithm. The first one tightens min-cut bound, one of two bounding strategies. Another technique exploits a graph-based approximation to a support vector machine problem to avoid the most time-consuming step. The last one tries to fix the labels of unlabeled examples whose labels can be obviously predicted based on labeled examples. Experimental results are presented which demonstrate that the proposed techniques can reduce drastically the number of subproblems and eventually computational time.

• Journal of Korean Port Research
• /
• v.12 no.2
• /
• pp.291-302
• /
• 1998

A digital image processing technique which is able to be used for getting the velocity vector distribution of a surface of the spilt oil in the ocean without contacting the flow itself. This technique is based upon the PIV(Particle Imaging Velocimetry) technique and its system mainly consists of a high sensitive camera, a CCD camera, an image grabber, and a host computer in which an image processing algorithm is adopted for velocity vector acquisition. For the acquisition of the advective velocity vector of floating matters on the ocean, a new multi-frame tracking algorithm is proposed, and for the acquisition of the diffusion velocity vector distribution of the spilt oil onto the water surface, a high sensitive gray-level cross-correlation algorithm is proposed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.3
• /
• pp.816-839
• /
• 2023

In a random vector functional link (RVFL) network, shortcomings such as local optimal stagnation and decreased convergence performance cause a reduction in the accuracy of illumination correction by only inputting the weights and biases of hidden neurons. In this study, we proposed an improved regularized random vector functional link (RRVFL) network algorithm with an optimized grey wolf optimizer (GWO). Herein, we first proposed the moth-flame optimization (MFO) algorithm to provide a set of excellent initial populations to improve the convergence rate of GWO. Thereafter, the MFO-GWO algorithm simultaneously optimized the input feature, input weight, hidden node and bias of RRVFL, thereby avoiding local optimal stagnation. Finally, the MFO-GWO-RRVFL algorithm was applied to ameliorate the performance of illumination correction of various test images. The experimental results revealed that the MFO-GWO-RRVFL algorithm was stable, compatible, and exhibited a fast convergence rate.

• Journal of the Korean Society of Industry Convergence
• /
• v.6 no.1
• /
• pp.3-10
• /
• 2003

Application of matrix converter to vector control of induction motor using simplified Venturini algorithm which is capable of achieving the maximum output voltage is developed. This algorithm simplifies the control algorithm and therefor reduces the digital implementation time. Matrix converter is used as voltage-referenced voltage fed vector controlled induction motor drive. This paper describes the performance of vector controlled induction motor with four quadrant capability employing a matrix converter power circuit. The advantage of this system over the conventional rectifier-inverter arrangement are capability for regeneration into the utility, sinusoidal supply currents and minimum passive components. The steady-state and transient performance of the induction motor drive under the vector control technique is demonstrate with simulation and experiment results.

• Journal of information and communication convergence engineering
• /
• v.16 no.3
• /
• pp.142-147
• /
• 2018

We present a practical design algorithm for quantizers at nodes in distributed systems in which each local measurement is quantized without communication between nodes and transmitted to a fusion node that conducts estimation of the parameter of interest. The benefits of vector quantization (VQ) motivate us to incorporate the VQ strategy into our design and we propose a low-complexity design technique that seeks to assign vector codewords into sets such that each codeword in the sets should be closest to its associated local codeword. In doing so, we introduce new distance metrics to measure the distance between vector codewords and local ones and construct the sets of vector codewords at each node to minimize the average distance, resulting in an efficient and independent encoding of the vector codewords. Through extensive experiments, we show that the proposed algorithm can maintain comparable performance with a substantially reduced design complexity.

• Journal of Biomedical Engineering Research
• /
• v.31 no.6
• /
• pp.456-463
• /
• 2010

In the present study, we proposed a new subspace scanning algorithm to enhance the spatial resolution of electroencephalography (EEG) and magnetoencephalography(MEG) source localization. Subspace scanning algorithms, represented by the multiple signal classification (MUSIC) algorithm and the first principal vector (FINE) algorithm, have been widely used to localize asynchronous multiple dipolar sources in human cerebral cortex. The conventional MUSIC algorithm used principal component analysis (PCA) to extract the noise vector subspace, thereby having difficulty in discriminating two or more closely-spaced cortical sources. The FINE algorithm addressed the problem by using only a part of the noise vector subspace, but there was no golden rule to determine the number of noise vectors. In the present work, we estimated a non-orthogonal signal vector set using independent component analysis (ICA) instead of using PCA and performed the source scanning process in the signal vector subspace, not in the noise vector subspace. Realistic 2D and 3D computer simulations, which compared the spatial resolutions of various algorithms under different noise levels, showed that the proposed ICA-MUSIC algorithm has the highest spatial resolution, suggesting that it can be a useful tool for practical EEG/MEG source localization.

• The KIPS Transactions:PartB
• /
• v.8B no.5
• /
• pp.539-548
• /
• 2001

In this paper, we propose that fast tracking algorithm for moving object is separated from background, using partial boundary line information. After detecting boundary line from input image, we track moving object by using the algorithm which takes boundary line information as feature of moving object. we extract moving vector on the imput image which has environmental variation, using high-performance BMA, and we extract moving object on the basis of moving vector. Next, we extract boundary line on the moving object as an initial feature-vector generating step for the moving object. Among those boundary lines, we consider a part of the boundary line in every direction as feature vector. And then, as a step for the moving object, we extract moving vector from feature vector generated under the information of the boundary line of the moving object on the previous frame, and we perform tracking moving object from the current frame. As a result, we show that the proposed algorithm using feature vector generated by each directional boundary line is simple tracking operation cost compared with the previous active contour tracking algorithm that changes processing time by boundary line size of moving object. The simulation for proposed algorithm shows that BMA operation is reduced about 39% in real image and tracking error is less than 2 pixel when the size of feature vector is [$10{\times}5$] using the information of each direction boundary line. Also the proposed algorithm just needs 200 times of search operation bout processing cost is varies by the size of boundary line on the previous algorithm.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.25 no.2
• /
• pp.211-219
• /
• 1988

In this paper effective block matching algorithms are proposed to find the motion vector. There are two approaches to the estimation of the motion vector in MCC (motion compensated coding), i.e.pel(pixel element) recursive algorithm and block matching algorithm. The search algorithm in this paper is based on the block matching method. The advantage of this algorithm is the reduction of the computation time. In order to reduce the computation time, three mathods are proposed in this paper. These new algorithms are faster than other methods. Compared with the three step algorithm by Koga et al., the average ratio of the computational savings obtained from the proposed algorithm is about 3-4.

• Journal of Korea Multimedia Society
• /
• v.13 no.10
• /
• pp.1487-1493
• /
• 2010

H.264 processes more detailed and more motion information for the compression efficiency. However, motion vector of H.264 takes more portion than previous standards such as MPEG-1/2/4 do, so that it is needed to consider motion vectors. This paper proposes the new algorithm with leaf mode in order to compress the motion vector efficiently. The proposed algorithm concentrates modes' distribution with leaf mode and carries out the compression of motion vector with less modes. The proposed algorithm adopted in current $4{\times}4$ motion vector matrix also can be extend to $8{\times}8$. The experiments shows that the proposed algorithm reduces up to 12.68% at header and 9.7% at resultant bitstream.

• Proceedings of the KIEE Conference
• /
• 1989.07a
• /
• pp.189-194
• /
• 1989

The sparse vector method is more efficient than conventional sparse matrix method when solving sparse system. This paper considers the structural relation between factorized L and inverse of L and presents a new ordering algorithm for sparse vector method. The method is useful in enhancing the sparsity of the inverse of L while preserving the aparsity of matrix. The performance of algorithm is compared with conventional algorithms by means of several power system.