• Smart Media Journal
• /
• v.11 no.3
• /
• pp.26-30
• /
• 2022

Spiking neural network is a neural network that applies the working principle of real brain neurons. Due to the biological mechanism of neurons, it consumes less power for training and reasoning than conventional neural networks. Recently, as deep learning models become huge and operating costs increase exponentially, the spiking neural network is attracting attention as a third-generation neural network that connects convolution neural networks and recurrent neural networks, and related research is being actively conducted. However, in order to apply the spiking neural network model to the industry, a lot of research still needs to be done, and the problem of model retraining to apply a new model must also be solved. In this paper, we propose a method to minimize the cost of model retraining by extracting the weights of the existing trained deep learning model and converting them into the weights of the spiking neural network model. In addition, it was found that weight conversion worked correctly by comparing the results of inference using the converted weights with the results of the existing model.

• Journal of Intelligence and Information Systems
• /
• v.4 no.1
• /
• pp.133-147
• /
• 1998

Recently, many researchers have been involved in finding deterministic equations which can accurately predict future event, based on chaotic theory, or fractal theory. The theory says that some events which seem very random but internally deterministic can be accurately predicted by fractal equations. In contrast to the conventional methods, such as AR model, MA, model, or ARIMA model, the fractal equation attempts to discover a deterministic order inherent in time series data set. In discovering deterministic order, researchers have found that neural networks are much more effective than the conventional statistical models. Even though prediction accuracy of the network can be different depending on the topological structure and modification of the algorithms, many researchers asserted that the neural network systems outperforms other systems, because of non-linear behaviour of the network models, mechanisms of massive parallel processing, generalization capability based on adaptive learning. However, recent survey shows that prediction accuracy of the forecasting models can be determined by the model structure and data structures. In the experiments based on actual economic data sets, it was found that the prediction accuracy of the neural network model is similar to the performance level of the conventional forecasting model. Especially, for the data set which is deterministically chaotic, the AR model, a conventional statistical model, was not significantly different from the MLP model, a neural network model. This result shows that the forecasting model. This result shows that the forecasting model a, pp.opriate to a prediction task should be selected based on characteristics of the time series data set. Analysis of the characteristics of the data set was performed by fractal analysis, measurement of Hurst index, and measurement of Lyapunov exponents. As a conclusion, a significant difference was not found in forecasting future events for the time series data which is deterministically chaotic, between a conventional forecasting model and a typical neural network model.

• Journal of Korean Society of Transportation
• /
• v.10 no.2
• /
• pp.25-42
• /
• 1992

The purpose of this paper is to construct a model to compute a progression adjustment factor on a signalized network. In a way to construct the model, a simulation method is introduced and the TRAF-NETSIM is used as a tool of simulation. The structure of the network chooses an urban arterial network so as to measure the effect of progression and compute average stopped delay on each link. A regression model is constructed by using the results of the simulation. The stepwise variable selection in the regression model in used. The findings of this paper are as follows: i)The secondary queue and platoon ratio are sensitive to the values of the progression adjustment factor ii) The continuous model can practically reflect on various situations in the real world. The platoon adjustment factor can be computed by this model and the data required for this model can be easily obtained in the field.

• Journal of Electrical Engineering and Technology
• /
• v.11 no.2
• /
• pp.265-273
• /
• 2016

A water wall system is one of the most important components of a boiler in a thermal power plant, and it is a nonlinear Multi-Input and Multi-Output (MIMO) system, with 6 inputs and 3 outputs. Three models are developed and comp for the controller design, including a linear model, a multilayer feed-forward neural network (MFNN) model and an Echo State Network (ESN) model. First, the linear model is developed by linearizing a given nonlinear model and is analyzed as a function of the operating point. Second, the MFNN and the ESN are developed by using training data from the nonlinear model. The three models are validated using Matlab with nonlinear input-output data that was not used during training.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.8B
• /
• pp.529-538
• /
• 2007

In this paper, we have analyzed a simulation model of Slammer worm propagation process which caused serious disruptions on the Internet in the year of 2003 by using NS-2. Previously we had presented and analyzed Abstract Network to Abstract Network(AN-AN) model being modified from the Detailed Network to Abstract Network(DN-AN) of NS-2. However, packet analysis in AN-AN model had a problem of taking 240 hours to simulate the initial 300 seconds of infection. We have reduced the AN-AN model to save the simulation time and analyzed total 3.5 hours of the network congestions within 107 hours. Moreover, we have derived optimal quarantine rate of 0.0022 considering service outage of network devices caused by the heavy infected traffics, which was not taken into consideration in previous works. As the result of simulation, Although the inbound traffic at the Korean international gateway was back in normal conditions at 4,787 second, due to the revese direction saturation was maintained until 12,600 seconds, the service outage was persisted for 3.5 hours.

• International journal of advanced smart convergence
• /
• v.8 no.4
• /
• pp.169-178
• /
• 2019

Due to the characteristics of an ad hoc network without a control center, self-organization, and flexible topology, the trust evaluation of the nodes in the network is extremely difficult. Based on the analysis of ad hoc networks and the blockchain technology, a blockchain-based node-level trust evaluation model is proposed. The concepts of the node trust degree of the HASH list on the blockchain and the perfect reward and punishment mechanism are adopted to construct the node trust evaluation model of the ad hoc network. According to the needs of different applications the network security level can be dynamically adjusted through changes in the trust threshold. The simulation experiments demonstrate that ad-hoc on-demand distance vector(AODV) Routing protocol based on this model of multicast-AODV(MAODV) routing protocol shows a significant improvement in security compared with the traditional AODV and on-demand multipath distance vector(AOMDV) routing protocols.

• Management Science and Financial Engineering
• /
• v.6 no.2
• /
• pp.29-45
• /
• 2000

We consider the design of a two-level telecommunication network having logical full-mesh/star topology, with the implementation of conduit systems taken together. The design problem is then viewed as consisting of three subproblems: locating hub facilities, placing a conduit network, and installing cables therein to configure the logical full-mesh/star network. Without partitioning into subproblems as done in the conventional approach, the whole problem is directly dealt with in a single integrated framework, inspired by some recent successes with the approach. We successfully formulate the problem as a variant of the classical multicommodity flow model for the fixed charge network design problem, aided by network augmentation, judicious commodity definition, and some flow restrictions. With our optimal model, we solve some randomly generated sample problems by using CPLEX MIP program. From the computational experiments, it seems that our model can be applied to the practical problem effectively.

• The Journal of the Korea Contents Association
• /
• v.5 no.1
• /
• pp.169-178
• /
• 2005

This paper presented high-level function model to achieve traffic engineering to construct traffic engineering infrastructure in Internet. Function model presented include traffic management, capacity management, and network planing. It is ensured that network performance is maximized under all conditions including load shifts and failures by traffic management. It is ensured that the network is designed and provisioned to meet performance objectives for network demands at minimum cost by capacity management. Also it is ensured that node and transport capacity is planned and deployed in advance of forecasted traffic growth by network planning.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.1
• /
• pp.120-126
• /
• 1995

The diffusion neural network can be cfficiently applied to the Gaussian processing. For example, a difference of two Gaussians(DOG) is performed by this network with ease. In this paper, we model a neural network to perform the function /t(.del.${\Delta}^{2}$G) by using the diffusion neural network. This model is used to detect the edges of moving target in image. By this model not only moving target is separated from stationary background but also their trajectories are obtained using accumulated past information in the diffusion neural network. Furthermore this model needs a small number of connections per cell and the connection weights are fixed-valued. Therefore its hardware can be easily implemented with simple structure.

• Proceedings of the Korean Society for Agricultural Machinery Conference
• /
• 1996.06c
• /
• pp.423-433
• /
• 1996

This study was conducted to predict the drafts of various tillage tools with a model tool and a neural network. Drafts of tillage tools were measured and a time lagged recurrent neural network was developed. The neural network had a structure to predict dynamic draft, having a function of one step ahead prediction . The results showed the model tool draft had linear relations with high coefficient of determinations to the drafts of the tillage tools. Also, the drafts of tillage tools were successfully predicted by the developed neural network.