• Journal of Broadcast Engineering
• /
• v.25 no.1
• /
• pp.1-12
• /
• 2020

In the medical field, numerical imbalance of data due to differences in disease prevalence is a common problem. It reduces the performance of a artificial intelligence network, leading to difficulties in learning a network with good performance. Recently, generative adversarial network (GAN) technology has been introduced as a way to address this problem, and its ability has been demonstrated by successful applications in various fields. However, it is still difficult to achieve good results in solving problems with performance degraded by numerical imbalances because the image resolution of the previous studies is not yet good enough and the structure in the image is modeled locally. In this paper, we propose a multi-scale conditional generative adversarial network based on attention mechanism, which can produce high resolution images to solve the numerical imbalance problem of chest X-ray image data. The network was able to produce images for various diseases by controlling condition variables with only one network. It's efficient and effective in that the network don't need to be learned independently for all disease classes and solves the problem of long distance dependency in image generation with self-attention mechanism.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2007.06a
• /
• pp.401-405
• /
• 2007

In sensor network, cluster based routing protocol about efficient energy usage method has researched variously. But existing cluster based routing protocol have problems. one of the problem is sensor nodes's imbalance energy consumption problem at cluster reconstruction. anther is non- connection problem between header node and spc node when they are far from each other, not properly connected. We propose cluster re-division and header node of multihop transmission method in this paper. The cluster re-division method is the method that re-divides existing routing protocol with the small-scale cluster and multihop transmission method is the method regarding the relay transmission between the header nodes. Through the simulation, the proposed routing mechanism shows more excellent than exiting routing protocol in balance energy consumption and energy efficiency.

• ETRI Journal
• /
• v.33 no.6
• /
• pp.857-863
• /
• 2011

Due to uncertainties in target motion and randomness of deployed sensor nodes, the problem of imbalance of energy consumption arises from sensor scheduling. This paper presents an energy-efficient adaptive sensor scheduling for a target monitoring algorithm in a local monitoring region of wireless sensor networks. Owing to excessive scheduling of an individual node, one node with a high value generated by a decision function is preferentially selected as a tasking node to balance the local energy consumption of a dynamic clustering, and the node with the highest value is chosen as the cluster head. Others with lower ones are in reserve. In addition, an optimization problem is derived to satisfy the problem of sensor scheduling subject to the joint detection probability for tasking sensors. Particles of the target in particle filter algorithm are resampled for a higher tracking accuracy. Simulation results show this algorithm can improve the required tracking accuracy, and nodes are efficiently scheduled. Hence, there is a 41.67% savings in energy consumption.

• Information Systems Review
• /
• v.22 no.1
• /
• pp.59-72
• /
• 2020

Artificial Intelligence is establishing itself as a familiar tool from an intractable concept. In this trend, financial sector is also looking to improve the problem of existing system which includes Fraud Detection System (FDS). It is being difficult to detect sophisticated cyber financial fraud using original rule-based FDS. This is because diversification of payment environment and increasing number of electronic financial transactions has been emerged. In order to overcome present FDS, this paper suggests 3 types of artificial intelligence models, Generative Adversarial Network (GAN), Deep Neural Network (DNN), and Convolutional Neural Network (CNN). GAN proves how data imbalance problem can be developed while DNN and CNN show how abnormal financial trading patterns can be precisely detected. In conclusion, among the experiments on this paper, WGAN has the highest improvement effects on data imbalance problem. DNN model reflects more effects on fraud classification comparatively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.9
• /
• pp.4207-4222
• /
• 2016

In a P2P (Peer-to-Peer) VoD (video-on-Demand) streaming system, the nodes' load is an important factor which affects the system performance. In the system, some nodes may receive too many requests, which leads to overload. On the other hand, some other nodes may receive too few requests, which leads to low utilization. Therefore, designing a reasonable load balancing strategy is important. However, existing related studies cannot handle this problem effectively, because they don't have an efficient dynamic load information management mechanism, and they don't distinguish the difference of requests when transfer the nodes' load. In this paper, to manage the dynamic load information efficiently, we design a load management table for each node. Based on the load information, we propose a load balancing strategy which uses a request migration algorithm (LBRM). Through simulations, our scheme can handle the load imbalance problem effectively and improve the users' playback fluency.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.10
• /
• pp.5143-5158
• /
• 2018

Improving the intrusion detection system (IDS) is a pressing need for cyber security world. With the growth of computer networks, there are constantly daily new attacks. Machine Learning (ML) is one of the most important fields which have great contribution to address the intrusion detection issues. One of these issues relates to the imbalance of the diverse classes of network traffic. Accuracy paradox is a result of training ML algorithm with imbalanced classes. Most of the previous efforts concern improving the overall accuracy of these models which is truly important. However, even they improved the total accuracy of the system; it fell in the accuracy paradox. The seriousness of the threat caused by the minor classes and the pitfalls of the previous efforts to address this issue is the motive for this work. In this paper, we consolidated stratified sampling, cost function and weighted Support Vector Machine (WSVM) method to address the accuracy paradox of ID problem. This model achieved good results of total accuracy and superior results in the small classes like the User-To-Remote and Remote-To-Local attacks using the improved version of the benchmark dataset KDDCup99 which is called NSL-KDD.

• Journal of Power Electronics
• /
• v.12 no.2
• /
• pp.344-356
• /
• 2012

In recent years, multilevel technology has become an effective and practical solution in the field of moderate and high voltage applications. This paper discusses an APF with a three-level NPC inverter. Obviously, the application of such converter to APFs is hindered by the problem of the voltage unbalance of DC capacitors, which leads to system instability. This paper comprehensively analyzes the theoretical limitations of the neutral-point voltage balancing problem for tracking different harmonic currents utilizing current switching functions from the space vector PWM (SVPWM) point of view. The fluctuation of the neutral point caused by the load currents of certain order harmonic frequency is reported and quantified. Furthermore, this paper presents a close-loop digital control algorithm of the DC voltage for this APF. A PI controller regulates the DC voltage in the outer-loop controller. In the current-loop controller, this paper proposes a simple neutral-point voltage control method. The neutral-point voltage imbalance is restrained by selecting small vectors that will move the neutral-point voltage in the direction opposite the direction of the unbalance. The experiment results illustrate that the performance of the proposed approach is satisfactory.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4144-4165
• /
• 2018

Software-defined networking (SDN) has emerged as a promising technology for network programmability and experiments. In this work, we focus on virtual network embedding in multiple controllers SDN network. In SDN virtualization environment, virtual SDN networks (vSDNs) operate on the shared substrate network and managed by their each controller, the placement and load of controllers affect vSDN embedding process. We consider controller placement, vSDN embedding, controller adjustment as a joint problem, together considering different quality of service (QoS) requirement for users, formulate the problem into mathematical models to minimize the average time delay of control paths, the load imbalance degree of controllers and embedding cost. We propose a heuristic method which places controllers and partitions control domains according to substrate SDN network, embeds different QoS constraint vSDN requests by corresponding algorithms, and migrates switches between control domains to realize load balance of controllers. The simulation results show that the proposed method can satisfy different QoS requirement of tenants, keep load balance between controllers, and work well in the acceptance ratio and revenue to cost ratio for vSDN embedding.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.7
• /
• pp.11-17
• /
• 2010

This paper proposes an improved SMO solving a quadratic optmization problem for class imbalanced learning. The SMO algorithm is aproporiate for solving the optimization problem of a support vector machine that assigns the different regularization values to the two classes, and the prosoposed SMO learning algorithm iterates the learning steps to find the current optimal solutions of only two Lagrange variables selected per class. The proposed algorithm is tested with the UCI benchmarking problems and compared to the experimental results of the SMO algorithm with the g-mean measure that considers class imbalanced distribution for gerneralization performance. In comparison to the SMO algorithm, the proposed algorithm is effective to improve the prediction rate of the minority class data and could shorthen the training time.

• Korean System Dynamics Review
• /
• v.6 no.1
• /
• pp.125-145
• /
• 2005

The Sihwaho Project got off to a bad start, which led to a series of mishaps and an imbalance of the whole project. The purpose of this study is to select the case of Sihwaho as a research subject, clean up the contamination caused by the ill planned project, develop suitable measures to stabilize the lake environment, and find the implications of similar development projects. For this, the authorshave conducted three simulations after studying the structural reasons for the failure of the Sihwaho Policy by identifying cause and effect relationships, pre-testing a number of policy measures for the current lake project, and presenting measures for solving the contamination problem at the lake. The simulations have shown us that filling the lake with seawater is inadequate to solve the problem and that we also have to make efforts to reduce the volume of wastes to the lake as well. The authorshave also analyzed the degree of difference between the simulation and survey results for scenario 1, in which we have studied how much seawater can reduce the contamination of the Sihwaho Lake without the effort to reduce the volume of wastes into the lake. The survey showed that most citizens and employees of the Ministry of Environment did not think it would be serious as the simulation results pointed out, and the employees of the Ministry of Environment were more optimistic about the situation than the public.