• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.620-626
• /
• 2004

For general nonlinear processes, it is difficult to control with a linear model-based control method and nonlinear controls are considered. Among the numerous approaches suggested, the most rigorous approach is to use dynamic optimization. Many general engineering problems like control, scheduling, planning etc. are expressed by functional optimization problem and most of them can be changed into dynamic programming (DP) problems. However the DP problems are used in just few cases because as the size of the problem grows, the dynamic programming approach is suffered from the burden of calculation which is called as 'curse of dimensionality'. In order to avoid this problem, the Neuro-Dynamic Programming (NDP) approach is proposed by Bertsekas and Tsitsiklis (1996). To get the solution of seriously nonlinear process control, the interest in NDP approach is enlarged and NDP algorithm is applied to diverse areas such as retailing, finance, inventory management, communication networks, etc. and it has been extended to chemical engineering parts. In the NDP approach, we select the optimal control input policy to minimize the value of cost which is calculated by the sum of current stage cost and future stages cost starting from the next state. The cost value is related with a weight square sum of error and input movement. During the calculation of optimal input policy, if the approximate cost function by using simulation data is utilized with Bellman iteration, the burden of calculation can be relieved and the curse of dimensionality problem of DP can be overcome. It is very important issue how to construct the cost-to-go function which has a good approximate performance. The neural network is one of the eager learning methods and it works as a global approximator to cost-to-go function. In this algorithm, the training of neural network is important and difficult part, and it gives significant effect on the performance of control. To avoid the difficulty in neural network training, the lazy learning method like k-nearest neighbor method can be exploited. The training is unnecessary for this method but requires more computation time and greater data storage. The pH neutralization process has long been taken as a representative benchmark problem of nonlin ar chemical process control due to its nonlinearity and time-varying nature. In this study, the NDP algorithm was applied to pH neutralization process. At first, the pH neutralization process control to use NDP algorithm was performed through simulations with various approximators. The global and local approximators are used for NDP calculation. After that, the verification of NDP in real system was made by pH neutralization experiment. The control results by NDP algorithm was compared with those by the PI controller which is traditionally used, in both simulations and experiments. From the comparison of results, the control by NDP algorithm showed faster and better control performance than PI controller. In addition to that, the control by NDP algorithm showed the good results when it applied to the cases with disturbances and multiple set point changes.

• Journal of Korea Multimedia Society
• /
• v.21 no.2
• /
• pp.98-107
• /
• 2018

In this paper, we propose a method to classify the skeletal maturity with a small amount of hand wrist X-ray image using deep learning-based meta-learning. General deep-learning techniques require large amounts of data, but in many cases, these data sets are not available for practical application. Lack of learning data is usually solved through transfer learning using pre-trained models with large data sets. However, transfer learning performance may be degraded due to over fitting for unknown new task with small data, which results in poor generalization capability. In addition, medical images require high cost resources such as a professional manpower and mcuh time to obtain labeled data. Therefore, in this paper, we use meta-learning that can classify using only a small amount of new data by pre-trained models trained with various learning tasks. First, we train the meta-model by using a separate data set composed of various learning tasks. The network learns to classify the bone maturity using the bone maturity data composed of the radiographs of the wrist. Then, we compare the results of the classification using the conventional learning algorithm with the results of the meta learning by the same number of learning data sets.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.173.5-173
• /
• 2001

Q-Learning is a most widely used reinforcement learning, which addresses the question of how an autonomous agent can learn to choose optimal actions to achieve its goal about any one problem. Q-Learning can acquire optimal control strategies from delayed rewards, even when the agent has no prior knowledge of the effects of its action in the environment. If agent has an ability using previous knowledge, then it is expected that the agent can speed up learning by interacting with environment. We present a novel reinforcement learning method using domain knowledge, which is represented by problem-independent features and their classifiers. Here neural network are implied as knowledge classifiers. To show that an agent using domain knowledge can have better performance than the agent with standard Q-Learner. Computer simulations are ...

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.120-123
• /
• 1997

This paper presents models of fuzzy inference systems that can be built from a set of input-output training data pairs through hybrid structure-parameter learning. Fuzzy inference systems has the difficulty of parameter learning. Here we develop a coding format to determine a fuzzy neural network(FNN) model by chromosome in a genetic algorithm(GA) and present systematic approach to identify the parameters and structure of FNN. The proposed FNN can automatically identify the fuzzy rules and tune the membership functions by modifying the connection weights of the networks using the GA and the back-propagation learning algorithm. In order to show effectiveness of it we simulate and compare with conventional methods.

• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10a
• /
• pp.870-876
• /
• 1993

This paper presents a binary clustering network (BCN) and a heuristic algorithm to detect pitch for recognition of keywords in continuous speech. In order to classify nonlinear patterns, BCN separates patterns into binary clusters hierarchically and links same patterns at root level by using the supervised learning and the unsupervised learning. BCN has many desirable properties such as flexibility of dynamic structure, high classification accuracy, short learning time, and short recall time. Pitch Detection algorithm is a heuristic model that can solve the difficulties such as scaling invariance, time warping, time-shift invariance, and redundance. This recognition algorithm has shown recognition rates as high as 95% for speaker-dependent as well as multispeaker-dependent tests.

• Proceedings of the KIEE Conference
• /
• 1998.07b
• /
• pp.578-580
• /
• 1998

The paper presents an approach to the structure identification based on the CART (Classification And Regression Tree) algorithm and to the parameter identification by hybrid learning method in neuro-fuzzy system. By using the CART algorithm, the proposed method can roughly estimate the numbers of membership function and fuzzy rule using the centers of decision regions. Then the parameter identification is carried out by the hybrid learning scheme using BP (Back-propagation) and RLSE (Recursive Least Square Estimation) from the numerical data. Finally, we will show it's usefulness for fuzzy modeling to truck backer upper control.

• Journal of Computing Science and Engineering
• /
• v.10 no.4
• /
• pp.111-117
• /
• 2016

An accurate approach for diagnosis of attention deficit hyperactivity disorder (ADHD) is presented in this paper. The presented technique efficiently classifies three subtypes of ADHD (ADHD-C, ADHD-H, ADHD-I) and typically developing control (TDC) by using only structural magnetic resonance imaging (MRI). The research examines structural MRI of the hippocampus from the ADHD-200 database. Each available MRI has been processed by a region-of-interest (ROI) to build a set of features for further analysis. The presented ADHD diagnostic approach unifies feature selection and classification techniques. The feature selection technique based on the proposed binary-coded genetic algorithm searches for an optimal subset of features extracted from the hippocampus. The classification technique uses a chosen optimal subset of features for accurate classification of three subtypes of ADHD and TDC. In this study, the famous Extreme Learning Machine is used as a classification technique. Experimental results clearly indicate that the presented BCGA-ELM (binary-coded genetic algorithm coupled with Extreme Learning Machine) efficiently classifies TDC and three subtypes of ADHD and outperforms existing techniques.

• Journal of the Korean Institute of Navigation
• /
• v.23 no.3
• /
• pp.35-44
• /
• 1999

This research aims to seek active control of ball-beam position stability by resorting to neural networks whose layers are given bias weights. The controller consists of an LQR (linear quadratic regulator) controller and a neural networks controller in parallel. The latter is used to improve the responses of the established LQR control system, especially when controlling the system with nonlinear factors or modelling errors. For the learning of this control system, the feedback-error learning algorithm is utilized here. While the neural networks controller learns repetitive trajectories on line, feedback errors are back-propagated through neural networks. Convergence is made when the neural networks controller reversely learns and controls the plant. The goals of teaming are to expand the working range of the adaptive control system and to bridge errors owing to nonlinearity by adjusting parameters against the external disturbances and change of the nonlinear plant. The motion equation of the ball-beam system is derived from Newton's law. As the system is strongly nonlinear, lots of researchers have depended on classical systems to control it. Its applications of position control are seen in planes, ships, automobiles and so on. However, the research based on artificial control is quite recent. The current paper compares and analyzes simulation results by way of the LQR controller and the neural network controller in order to prove the efficiency of the neural networks control algorithm against any nonlinear system.

• 제어로봇시스템학회:학술대회논문집
• /
• 2002.10a
• /
• pp.120.2-120
• /
• 2002

In this paper auto-tuning scheme of weight function in the neural networks has been suggested by immune algorithm for nonlinear process. A number of structures of the neural networks are considered as learning methods for control system. A general view is provided that they are the special cases of either the membership functions or the modification of network structure in the neural networks. On the other hand, since the immune network system possesses a self organizing and distributed memory, it is thus adaptive to its external environment and allows a PDP (parallel distributed processing) network to complete patterns against the environmental situation. Also. It can provi..

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10a
• /
• pp.878-883
• /
• 1991

In this paper, we propose the fuzzy-neural controller which is fuzzy controller with learning ability of neural network. The neural network in this controller is same as the membership function in current fuzzy controller and a parts of inference rules. And, it can be easily extend the control algorithm to multivariable systems. We can show effectiveness of the control algorithm through experiment of the inverted pendulum system.