• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.2
• /
• pp.161-168
• /
• 2024

As the size of big data models grows, distributed training is emerging as an essential element for large-scale machine learning tasks. In this paper, we propose ParamHub for distributed data training. During the training process, this agent utilizes the provided data to adjust various conditions of the model's parameters, such as the model structure, learning algorithm, hyperparameters, and bias, aiming to minimize the error between the model's predictions and the actual values. Furthermore, it operates autonomously, collecting and updating data in a distributed environment, thereby reducing the burden of load balancing that occurs in a centralized system. And Through communication between agents, resource management and learning processes can be coordinated, enabling efficient management of distributed data and resources. This approach enhances the scalability and stability of distributed machine learning systems while providing flexibility to be applied in various learning environments.

• ETRI Journal
• /
• v.36 no.5
• /
• pp.741-751
• /
• 2014

Smart TV is expected to bring cloud services based on virtualization technologies to the home environment with hardware and software support. Although most physical resources can be shared among virtual machines (VMs) using a time sharing approach, allocating the proper amount of memory to VMs is still challenging. In this paper, we propose a novel mechanism to dynamically balance the memory allocation among VMs in virtualized Smart TV systems. In contrast to previous studies, where a virtual machine monitor (VMM) is solely responsible for estimating the working set size, our mechanism is symbiotic. Each VM periodically reports its memory usage pattern to the VMM. The VMM then predicts the future memory demand of each VM and rebalances the memory allocation among the VMs when necessary. Experimental results show that our mechanism improves performance by up to 18.28 times and reduces expensive memory swapping by up to 99.73% with negligible overheads (0.05% on average).

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.05a
• /
• pp.529-532
• /
• 2012

This paper presents the system modeling, analysis, and controller design and implementation with a inverted pendulum system in order to test robust algorithm for sweeping machine. The balancing of an inverted pendulum by moving pendulum robot like as 'segway' along a horizontal track is a classic problem in the area of control. This paper will describe two methods to swing a pendulum attached to a cart from an initial downwards position to an upright position and maintain that state. The results of real experiment show that the proposed control system has superior performance for following a reference command at certain initial conditions.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.06a
• /
• pp.280-283
• /
• 2005

This study performed by a previous research for the applying expert system to active vibration control algorithm. In order to increase productivity and efficiency, high-speed rotating machines become popular these days. They are likely to vibrate and cause machine failure even though they have small unbalance. Therefore, a high-speed rotating machine needs a balancing technique. ISO 11342 classifies flexible rotors in accordance with their balancing requirements and establishes methods of assessment of residual unbalance. But, even if they finished balancing work, they have harmful effect vibration under the high-speed rotating environment. This vibration effect is very small, but it must be removed for the improvement of the rotor's spin accuracy. This paper introduces a new active control method that remove the exciting force by a phase control. For this method, the high-speed rotating rotor was reconstructed by a flexible rotor model. The forces which excite the rotating system suppose cyclic forces, we obtain the responses by numerical method. And then through the pattern analysis about the vibraton responses, the controler generate the control force with the reverse phase and similar magnitude. This paper suggest an phase control method and shows how to improve the rotating vibration accuracy of the flexible rotor dynamics system using phase control method.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4103-4121
• /
• 2018

Quality of Service (QoS) awareness is recognized as a key point for the success of Internet of Things (IOT).Realizing the full potential of the Internet of Things requires, a real-time task scheduling algorithm must be designed to meet the QoS need. In order to schedule tasks with diverse QoS requirements in cloud environment efficiently, we propose a task scheduling strategy based on dynamic priority and load balancing (DPLB) in this paper. The dynamic priority consisted of task value density and the urgency of the task execution, the priority is increased over time to insure that each task can be implemented in time. The scheduling decision variable is composed of time attractiveness considered earliest completion time (ECT) and load brightness considered load status information which by obtain from each virtual machine by topic-based publish/subscribe mechanism. Then sorting tasks by priority and first schedule the task with highest priority to the virtual machine in feasible VMs group which satisfy the QoS requirements of task with maximal. Finally, after this patch tasks are scheduled over, the task migration manager will start work to reduce the load balancing degree.The experimental results show that, compared with the Min-Min, Max-Min, WRR, GAs, and HBB-LB algorithm, the DPLB is more effective, it reduces the Makespan, balances the load of VMs, augments the success completed ratio of tasks before deadline and raises the profit of cloud service per second.

• Journal of Information Processing Systems
• /
• v.7 no.1
• /
• pp.53-62
• /
• 2011

With the steady growth of mobile technology and applications, demand for more storage in mobile devices has also increased. A lightweight block-level protocol, Internet Advanced Technology Attachment (iATA), has been developed to deliver a cost-effective storage network solution for mobile devices to obtain more storage. This paper seeks to contribute to designing and implementing Load Balancing (LB), Network Monitoring (NM) and Write Replication (WR) modules to improve the protocol's scalability and data availability. LB and NM modules are invoked to collect system resources states and current network status at each associate node (server machine). A dynamic weight factor is calculated based on the collected information and sent to a referral server. The referral server is responsible to analyze and allocate the most ideal node with the least weight to serve the client. With this approach, the client can avoid connecting to a heavily loaded node that may cause delays in subsequent in-band I/O operations. Write replication is applied to the remaining nodes through a WR module by utilizing the Unison file synchronization program. A client initially connected to node IP A for write operations will have no hindrances in executing the relevant read operations at node IP B in new connections. In the worst case scenario of a node crashing, data remain recoverable from other functioning nodes. We have conducted several benchmark tests and our results are evaluated and verified in a later section.

• Journal of Information Processing Systems
• /
• v.16 no.1
• /
• pp.132-144
• /
• 2020

Cloud computing is an emerging technology based on the concept of enabling data access from anywhere, at any time, from any platform. The exponential growth of cloud users has resulted in the emergence of multiple issues, such as the workload imbalance between the virtual machines (VMs) of data centers in a cloud environment greatly impacting its overall performance. Our axis of research is the load balancing of a data center's VMs. It aims at reducing the degree of a load's imbalance between those VMs so that a better resource utilization will be provided, thus ensuring a greater quality of service. Our article focuses on two phases to balance the workload between the VMs. The first step will be the determination of the threshold of each VM before it can be considered overloaded. The second step will be a task allocation to the VMs by relying on an improved and faster version of the meta-heuristic "simulated annealing (SA)". We mainly focused on the acceptance probability of the SA, as, by modifying the content of the acceptance probability, we could ensure that the SA was able to offer a smart task distribution between the VMs in fewer loops than a classical usage of the SA.

• Journal of Biosystems Engineering
• /
• v.32 no.3
• /
• pp.160-165
• /
• 2007

This study was conducted to evaluate beef using a color machine vision system. The machine vision system has an advantage to measure larger area than a colorimeter and also could measure other quality factors like distribution of fats. However, the machine vision measurement is affected by system components. To measure the beef color with the machine vision system, the effect of color balancing control was tested and calibration model was developed. Neural network for color calibration which learned reference color patches showed a high correlation with colorimeter in L*a*b* coordinates and had an adaptability at various measurement environments. The trained network showed a very high correlation with the colorimeter when measuring beef color.

• Agricultural and Biosystems Engineering
• /
• v.4 no.1
• /
• pp.16-21
• /
• 2003

In the agricultural field, a machine vision system has been widely used to automate most inspection processes especially in quality grading. Though machine vision system was very effective in quantifying geometrical quality factors, it had a deficiency in quantifying color information. This study was conducted to evaluate color of beef using machine vision system. Though measuring color of a beef using machine vision system had an advantage of covering whole lean tissue area at a time compared to a colorimeter, it revealed the problem of sensitivity depending on the system components such as types of camera, lighting conditions, and so on. The effect of color balancing control of a camera was investigated and multi-layer BP neural network based color calibration process was developed. Color calibration network model was trained using reference color patches and showed the high correlation with L*a*b* coordinates of a colorimeter. The proposed calibration process showed the successful adaptability to various measurement environments such as different types of cameras and light sources. Compared results with the proposed calibration process and MLR based calibration were also presented. Color calibration network was also successfully applied to measure the color of the beef. However, it was suggested that reflectance properties of reference materials for calibration and test materials should be considered to achieve more accurate color measurement.

• Geomechanics and Engineering
• /
• v.37 no.6
• /
• pp.539-554
• /
• 2024

This study explores development of prediction model for seismic site classification through the integration of machine learning techniques with horizontal-to-vertical spectral ratio (HVSR) methodologies. To improve model accuracy, the research employs outlier detection methods and, synthetic minority over-sampling technique (SMOTE) for data balance, and evaluates using seven machine learning models using seismic data from KiK-net. Notably, light gradient boosting method (LGBM), gradient boosting, and decision tree models exhibit improved performance when coupled with SMOTE, while Multiple linear regression (MLR) and Support vector machine (SVM) models show reduced efficacy. Outlier detection techniques significantly enhance accuracy, particularly for LGBM, gradient boosting, and voting boosting. The ensemble of LGBM with the isolation forest and SMOTE achieves the highest accuracy of 0.91, with LGBM and local outlier factor yielding the highest F1-score of 0.79. Consistently outperforming other models, LGBM proves most efficient for seismic site classification when supported by appropriate preprocessing procedures. These findings show the significance of outlier detection and data balancing for precise seismic soil classification prediction, offering insights and highlighting the potential of machine learning in optimizing site classification accuracy.