• Journal of the Korean Institute of Intelligent Systems
• /
• v.10 no.5
• /
• pp.442-448
• /
• 2000

The conventional control of dynamic systems needs accurate mathematical modeling of control systems. But the modeling of dynamic systems require very complex computation process due to complex state equation and many control parameters. Accordingly this paper proposes a state space identification model of the dynamic system using neural networks. The Gauss-Newton method is used to train the proposed neural network and the effectiveness of proposed method is verified through the computer simulation of the Seesaw system identification problem.

• Proceedings of the IEEK Conference
• /
• 2002.07b
• /
• pp.1188-1191
• /
• 2002

A number of studies have recently been published concerning neuron models and asynchronous neural networks. In the case of large-scale neural networks having neuron models, the neural network should be constructed using analog hardware, rather than by computer simulation via software, because of the limitation of the computational power, In this paper, we discuss the circuit structure of a synaptic section model having the spatio-temporal summation of inputs and utilizing CMOS processing.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10a
• /
• pp.236-242
• /
• 1992

Multi-layered neural networks are applied to the identification and control of nonlinear dynamical system. Traditional adaptive control techniques can only deal with linear systems or some special nonlinear systems. A scheme for combining multi-layered neural networks with model reference network techniques has the capability to learn the nonlinearity and shows the great potential for adaptive control. In many interesting cases the system can be described by a nonlinear model in which the control input appears linearly. In this paper the identification of linear and nonlinear part are performed simultaneously. The projection algorithm and the new estimation method which uses the delta rule of neural network are compared throughout the simulation. The simulation results show that the identification and adaptive control schemes suggested are practically feasible and effective.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.3
• /
• pp.48-56
• /
• 1995

This paper presents a method of implementing efficient optical flow estimation for dynamic scene analysis using the hierarchical Hopfield neural networks. Given the two consequent inages, Zhou and Chellappa suggested the Hopfield neural network for computing the optical flow. The major problem of this algorithm is that Zhou and Chellappa's network accompanies self-feedback term, which forces them to check the energy change every iteration and only to accept the case where the lower the energy level is guaranteed. This is not only undesirable but also inefficient in implementing the Hopfield network. The another problem is that this model cannot allow the exact computation of optical flow in the case that the disparities of the moving objects are large. This paper improves the Zhou and Chellapa's problems by modifying the structure of the network to satisfy the convergence condition of the Hopfield model and suggesting the hierarchical algorithm, which enables the computation of the optical flow using the hierarchical structure even in the presence of large disparities.

• Proceedings of the KIEE Conference
• /
• 1998.07b
• /
• pp.566-568
• /
• 1998

In this paper, we suggest a optimal design method of Fuzzy-Neural Networks model for complex and nonlinear systems. FNNs have the stucture of fusion of both fuzzy inference with linguistic variables and Neural Networks. The network structure uses the simpified inference as fuzzy inference system and the BP algorithm as learning procedure. And we use a clustering algorithm to find initial parameters of membership function. The parameters such as membership functions, learning rates and momentum coefficients are easily adjusted using the genetic algorithms. Also, the performance index with weighted value is introduced to achieve a meaningful balance between approximation and generalization abilities of the model. To evaluate the performance index, we use the time series data for gas furnace and the sewage treatment process.

• Structural Engineering and Mechanics
• /
• v.44 no.3
• /
• pp.325-337
• /
• 2012

In this article, the multi-objective optimization of cylindrical aluminum tubes under axial impact load is presented. The specific absorbed energy and the maximum crushing force are considered as objective functions. The geometric dimensions of tubes including diameter, length and thickness are chosen as design variables. D/t and L/D ratios are constricted in the range of which collapsing of tubes occurs in concertina or diamond mode. The Non-dominated Sorting Genetic Algorithm-II is applied to obtain the Pareto optimal solutions. A back-propagation neural network is constructed as the surrogate model to formulate the mapping between the design variables and the objective functions. The finite element software ABAQUS/Explicit is used to generate the training and test sets for the artificial neural networks. To validate the results of finite element model, several impact tests are carried out using drop hammer testing machine.

• Proceedings of the KIEE Conference
• /
• 1993.07a
• /
• pp.221-223
• /
• 1993

The Hopfield model is defined as an adaptive dynamic system. In this paper we propose a modified neural network which is capable of solving linear programming problems and a set of linear equations. The model is directly implemented from the given system, and solves the problem without calculating the inverse of the matrices. We get the better stability results by the addition of scaling property and by using the nonlinearities in the linear programming neural networks.

• Geomechanics and Engineering
• /
• v.1 no.4
• /
• pp.307-321
• /
• 2009

The paper employs a feed forward neural network with back-propagation algorithm for modeling time dependent swell in clays containing carbonate in the presence of sulfuric acid. The oedometer swell percent is estimated at a nominal surcharge pressure of 6.25 kPa to develop 612 data sets for modeling. The input parameters used in the network include time, sulfuric acid concentration, carbonate percentage, and liquid limit. Among the total data sets, 280 (46%) were assigned to training, 175 (29%) for testing and the remaining 157 data sets (25%) were relegated to cross validation. The network was programmed to process this information and predict the percent swell at any time, knowing the variable involved. The study demonstrates that it is possible to develop a general BPNN model that can predict time dependent swell with relatively high accuracy with observed data ($R^2$=0.9986). The obtained results are also compared with generated non-linear regression model.

• Journal of Sensor Science and Technology
• /
• v.30 no.5
• /
• pp.286-289
• /
• 2021

Image classification is one of the fundamental applications of computer vision. It enables a system to identify an object in an image. Recently, image classification applications have broadened their scope from computer applications to edge devices. The convolutional neural network (CNN) is the main class of deep learning neural networks that are widely used in computer tasks, and it delivers high accuracy. However, CNN algorithms use a large number of parameters and incur high computational costs, which hinder their implementation in edge hardware devices. To address this issue, this paper proposes a lightweight image classifier that provides good accuracy while using fewer parameters. The proposed image classifier diverts the input into three paths and utilizes different scales of receptive fields to extract more feature maps while using fewer parameters at the time of training. This results in the development of a model of small size. This model is tested on the CIFAR-10 dataset and achieves an accuracy of 90% using .26M parameters. This is better than the state-of-the-art models, and it can be implemented on edge devices.

• Journal of Information Processing Systems
• /
• v.14 no.6
• /
• pp.1318-1330
• /
• 2018

Video captioning refers to the process of extracting features from a video and generating video captions using the extracted features. This paper introduces a deep neural network model and its learning method for effective video captioning. In this study, visual features as well as semantic features, which effectively express the video, are also used. The visual features of the video are extracted using convolutional neural networks, such as C3D and ResNet, while the semantic features are extracted using a semantic feature extraction network proposed in this paper. Further, an attention-based caption generation network is proposed for effective generation of video captions using the extracted features. The performance and effectiveness of the proposed model is verified through various experiments using two large-scale video benchmarks such as the Microsoft Video Description (MSVD) and the Microsoft Research Video-To-Text (MSR-VTT).