• Journal of Korea Multimedia Society
• /
• v.21 no.4
• /
• pp.441-449
• /
• 2018

Deep learning has been used for various processing centered on image recognition. One core algorithms of the deep learning, convolutional neural network is an deep neural network that specialized in image recognition. In this paper, we use a convolutional neural network to classify forest insects and propose an optimization method. Experiments were carried out by combining two weight initialization and six parameter update methods. As a result, the Xavier-SGD method showed the highest performance with an accuracy of 82.53% in the 12 different combinations of experiments. Through this, the latest learning algorithms, which complement the disadvantages of the previous parameter update method, we conclude that it can not lead to higher performance than existing methods in all application environments.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.9
• /
• pp.117-123
• /
• 1996

In this paper, we present a new learning method for the fuzzy ARTMAP which is effective for the noisy input patterns. Conventional fuzzy ARTMAP employs only fuzzy AND operation between input vector and weight vector in learning both top-down and bottom-up weight vectors. This fuzzy AND operation causes excessive update of the weight vector in the noisy input environment. As a result, the number of spurious categories are increased and the recognition ratio is reduced. To solve these problems, we propose a new method in updating the weight vectors: the top-down weight vectors of the fuzzy ART system are updated using weighted average of the input vector and the weight vector itself, and the bottom-up weight vectors are updated using fuzzy AND operation between the updated top-down weitht vector and bottom-up weight vector itself. The weighted average prevents the excessive update of the weight vectors and the fuzzy AND operation renders the learning fast and stble. Simulation results show that the proposed method reduces the generation of spurious categories and increases the recognition ratio in the noisy input environment.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.5
• /
• pp.2723-2740
• /
• 2017

Steganographic schemes which are based on minimizing an additive distortion function defined the overall impacts after embedding as the sum of embedding costs for individual image element. However, mutual impacts during embedding are often ignored. In this paper, an adaptive JPEG steganographic method based on weight distribution for embedding costs is proposed. The method takes mutual impacts during embedding in consideration. Firstly, an analysis is made about the factors that affect embedding fluctuations among JPEG coefficients. Then the Distortion Update Strategy (DUS) of updating the distortion costs is proposed, enabling to dynamically update the embedding costs group by group. At last, a kind of adaptive JPEG steganographic algorithm is designed combining with the update strategy and well-known additive distortion function. The experimental result illustrates that the proposed algorithm gains a superior performance in the fight against the current state-of-the-art steganalyzers with high-dimensional features.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.3
• /
• pp.1377-1389
• /
• 2016

This paper presents a target model update scheme for the mean-shift tracking with background weighted histogram. In the scheme, the target candidate histogram is corrected by considering the back-projection weight of each pixel in the kernel after the best target candidate in the current frame image is chosen. In each frame, the target model is updated by the weighted average of the current target model and the corrected target candidate. We compared our target model update scheme with the previous ones by applying several test sequences. The experimental results showed that the object tracking accuracy was greatly improved by using the proposed scheme.

• Journal of Communications and Networks
• /
• v.9 no.4
• /
• pp.499-510
• /
• 2007

Shortest path tree(SPT) construction is essential in high performance routing in an interior network using link state protocols. When some links have new state values, SPTs may be rebuilt, but the total rebuilding of the SPT in a static way for a large computer network is not only computationally expensive, unnecessary modifications can cause routing table instability. This paper presents a new update algorithm, dynamic shortest path tree(DSPT) that is computationally economical and that maintains the unmodified nodes mostly from an old SPT to a new SPT. The proposed algorithm reduces redundancy using a dynamic update approach where an edge becomes the significant edge when it is extracted from a built edge list Q. The average number of significant edges are identified through probability analysis based on an arbitrary tree structure. An update derived from significant edges is more efficient because the DSPT algorithm neglect most other redundant edges that do not participate in the construction of a new SPT. Our complexity analysis and experimental results show that DSPT is faster than other known methods. It can also be extended to solve the SPT updating problem in a graph with negative weight edges.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.3C
• /
• pp.157-162
• /
• 2011

For blind equalization of communication channels, concurrent equalization is useful to improve convergence characteristics. However, the concurrent equalization will result in limited performance enhancement by continuing concurrent adaptation with two algorithms after the equalizer converges to steady-state. In this paper, to improve the convergence characteristics and steady-state performance of the concurrent equalization, proposed is a new concurrent equalization technique with variable step-size parameter and weight-based tap coefficients update. The proposed concurrent vsCMA+DD equalization calculates weight factors using error signals of the variable step-size CMA (vsCMA) and DD (decision-directed) algorithm, and then updates the two equalizers based on the weights respectively. The proposed method, first, improves the error performance of the CMA by the vsCMA, and enhances the steady-state performance as well as the convergence speed further by the weight-based tap coefficients update. The performance improvement by the proposed scheme is verified through simulations.

• Journal of Institute of Control, Robotics and Systems
• /
• v.2 no.3
• /
• pp.142-147
• /
• 1996

In this paper, we describe the algorithm which controls an unknown nonlinear system with multilayer neural network. The multilayer neural network can be used to approximate any continuous function to any desired degree of accuracy. With the former fact, we approximate unknown nonlinear function on the nonlinear system by using of multilayer neural network. The weight-update rule of multilayer neural network is derived to satisfy Lyapunov stability. The whole control system constitutes controller using feedback linearization method. The weight of neural network which is used to implement nonlinear function is updated by the derived update-rule. The proposed control algorithm is verified through computer simulation.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.9
• /
• pp.41-49
• /
• 1997

In this paper, we describe the algorithm which controls an unknown nonlinear system with disturbance a using multilayer neural network. The multilayer neural network can be used to approximate any continuous function to any desired degree of accuracy. With the former fact, we approximate an unknown nonlinear system by using of multilayer neural netowrk. WE include a disturbance among the modelling error, and the weight-update rule of multilayer neural network is derived to satisfy Laypunov stability. The whole control system constitutes controller using the feedback linearization method. The weight of neural network which is used to implement nonlinear function is updated by the derived update-rule. The proposed control algorithm is verified through computer simulation.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2002.05a
• /
• pp.509-513
• /
• 2002

This paper presents a semi-blind Minimum Mean Square Error (MMSE) beamforming adaptive algorithm used far OFDM/CDMA combined system. The proposed algorithm exploits the transmitting pilot signal in the initial period of the transmission to update the weight vector. Then it applies the blind adaptive period to update the weight vector, in which the pilot signal is no longer used. The derivation of the algorithm based on the Mean Square Error (MSE) criterion is also presented. Computer simulation is carried out to verify the performance of the proposed approach.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.9
• /
• pp.2041-2047
• /
• 2013

In this paper, we proposed high speed LDPC decoding algorithm based on DVB-S2 standard. For implementing the high speed LDPC decoder, HSS algorithm which reduce the iteration numbers without performance degradation is applied. In HSS algorithm, check node update units are update at the same time of bit node update. HSS can be accelerated to the decoding speed because it does not need to separate calculation of the bit nodes, However, check node calculation blocks need many clocks because of just one memory is used. Therefore, this paper proposed dc-split memory structure in order to reduced the delay and high speed decoder is possible. Finally, this paper presented maximum split memory and throughput for various coding rates in DVB-S2 standard.