• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.05a
• /
• pp.381-384
• /
• 2006

Self Organizing Map(SOM) is a powerful neural network model for unsupervised loaming. In many clustering works with exploratory data analysis, it has been popularly used. But it has a weakness which is the poorly theoretical base. A lot more researches for settling the problem have been published. Also, our paper proposes a method to overcome the drawback of SOM. As compared with the presented researches, our method has a different approach to solve the problem. So, a hybrid SOM is proposed in this paper. Using Monte Carlo computing, a hybrid SOM improves the performance of clustering. We verify the improved performance of a hybrid SOM according to the experimental results using UCI machine loaming repository. In addition to, the number of clusters is determined by our hybrid SOM.

• Proceedings of the Korea Society for Simulation Conference
• /
• 2001.10a
• /
• pp.321-324
• /
• 2001

The robot soccer simulation game is a dynamic multi-agent environment. In this paper we suggest a new reinforcement learning approach to each agent's dynamic positioning in such dynamic environment. Reinforcement Beaming is the machine learning in which an agent learns from indirect, delayed reward an optimal policy to choose sequences of actions that produce the greatest cumulative reward. Therefore the reinforcement loaming is different from supervised teaming in the sense that there is no presentation of input-output pairs as training examples. Furthermore, model-free reinforcement loaming algorithms like Q-learning do not require defining or loaming any models of the surrounding environment. Nevertheless it can learn the optimal policy if the agent can visit every state-action pair infinitely. However, the biggest problem of monolithic reinforcement learning is that its straightforward applications do not successfully scale up to more complex environments due to the intractable large space of states. In order to address this problem, we suggest Adaptive Mediation-based Modular Q-Learning(AMMQL) as an improvement of the existing Modular Q-Learning(MQL). While simple modular Q-learning combines the results from each learning module in a fixed way, AMMQL combines them in a more flexible way by assigning different weight to each module according to its contribution to rewards. Therefore in addition to resolving the problem of large state space effectively, AMMQL can show higher adaptability to environmental changes than pure MQL. This paper introduces the concept of AMMQL and presents details of its application into dynamic positioning of robot soccer agents.

• Journal of the Korean Society for Precision Engineering
• /
• v.15 no.2
• /
• pp.162-169
• /
• 1998

Lacking for the skilled grinding operator possessed of the experiential knowledges in machine shop, there is the just requirement which includes the establishment of the optimal grinding conditions. Accordingly, we attemt to develope the selection system of optimal grinding conditions such as workpiece velocity, depth of cut and wheel velocity and to add the trouble shooting system by means of the neural network. Those systems are robust to the each machine error and environmental unstable state. In addition. we produce the loaming process that is progressed with additional data modified by skilled operators, and excluding is advanced to similarity of input data.

• Journal of Educational Research in Mathematics
• /
• v.9 no.1
• /
• pp.263-278
• /
• 1999

Computer with various powerful functions has profound potential for mathematics instruction and learning. As computer technology progress, its applicability to mathematics education become more comprehensive. Not only its functional development but various psychological positions also changed the way computer technology utilized in mathematics education. In behaviorist's perspective, computer viewed as a teaching machine and constructivist viewed computer as microworld where students could explore various mathematical contents. Both theoretical positions emphasized individual aspect of learning because behaviorist tried to individualize learning using computer and constructivist focused on the process of individual construction. But learning is not only a individual event but also a social event. Therefore we must take social aspect into account. This is especially important when it comes to computer based learning. So far, mathematics loaming with computer weighed individual aspect of loaming. Even in microworld environment, learning should be mediated by teacher and collaborative learning activities. In this aspect, the roles of teacher and peers are very important and socio-cultural perspective sheds light on the computer based learning. In socio-cultural perspective, the idea of scaffold is very important in learning and students gradually internalize the social dimension and scaffolding is gradually faded. And in the zone of proximal development, teacher and more competent peers guide students to formulate their own understanding. In sum, we must take following points into account. First of all, computer should not be viewed as a medium for individualized teaming. That is, interaction with computer should be catalyst for collaborative activities with peers. So, exploration in computer environment has to be followed by small group activities including small group discussion. Secondly, regardless of the role that computer would play, teacher should play a crucial role in computer based learning. This does not mean teacher should direct every steps in learning process. Teacher's intervention should help student construct actively. Thirdly, it is needed to conceptualize computer in learning situation as medium. This would affect learning situation and result in the change of pre-service and in-service teacher training. Computer to be used effectively in mathematics classroom, researches on assessment of computer based learning are needed.

• 한국공간정보시스템학회:학술대회논문집
• /
• 2005.11a
• /
• pp.61-66
• /
• 2005

Spatiotemporal data mining represents the confluence of several fields including spatiotemporal databases, machine loaming, statistics, geographic visualization, and information theory. Exploration of spatial data mining and temporal data mining has received much attention independently in knowledge discovery in databases and data mining research community. In this paper, we introduce an algorithm Max_MOP for discovering moving sequences in mobile environment. Max_MOP mines only maximal frequent moving patterns. We exploit the characteristic of the problem domain, which is the spatiotemporal proximity between activities, to partition the spatiotemporal space. The task of finding moving sequences is to consider all temporally ordered combination of associations, which requires an intensive computation. However, exploiting the spatiotemporal proximity characteristic makes this task more cornputationally feasible. Our proposed technique is applicable to location-based services such as traffic service, tourist service, and location-aware advertising service.

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.11 no.4
• /
• pp.62-67
• /
• 2002

This paper presents a new approach to the design of cruise control system of a mobile robot with two drive wheel. The proposed control scheme uses a Gaussian function as a unit function in the fuzzy-neural network and back propagation algorithm to train the fuzzy-neural network controller in the framework of the specialized loaming architecture. It is Proposed a learning controller consisting of two neural network-fuzzy based on independent reasoning and a connection net with fixed weights to simply the neural networks-fuzzy. The performance of the proposed controller is shown by performing the computer simulation for trajectory tucking of the speed and azimuth of a mobile robot driven by two independent wheels.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2003.04a
• /
• pp.124-129
• /
• 2003

In the face of global competitor the requirements flor the continuously increasing productivity, flexibility and quality(dimensional accuracy, mechanical properties and surface properties) have imposed a major change on steel manufacturing industries. Indeed, one of the keys to achieve this goal is the automation of the steel-making process using AI(Artificial Intelligence) techniques. The automation of hot rolling process requires the developments of several mathematical models fir simulation and quantitative description of the industrial operations involved. In this paper, a on-line training neural network for both long-term teaming and short-term teaming was developed in order to improve the prediction of rolling force in hot rolling mill. This analysis shows that the predicted rolling force is very closed to the actual rolling force, and the thickness error of the strip is considerably reduced.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2003.04a
• /
• pp.305-311
• /
• 2003

We propose a new technique of adaptive-neuro controller design to implement real-time control of robot manipulator, Unlike the well-established theory for the adaptive control of linear systems, there exists relatively little general theory for the adaptive control of nonlinear systems. Adaptive control technique is essential for providing a stable and robust performance for application of robot control. The proposed neuro control algorithm is one of loaming a model based error back-propagation scheme using Lyapunov stability analysis method. Through simulation, the proposed adaptive-neuro control scheme is proved to be a efficient control technique for real time control of robot system using DSPs(TMS320C50)

• 제어로봇시스템학회:학술대회논문집
• /
• 1999.10a
• /
• pp.292-297
• /
• 1999

Evolving in artificial agent is an extremely difficult problem, but on the other hand, a challenging task. At present the studies mainly centered on single agent learning problem. In our case, we use simulated soccer to investigate multi-agent cooperative learning. Consider the fundamental differences in learning mechanism, existing reinforcement learning algorithms can be roughly classified into two types-that based on evaluation functions and that of searching policy space directly. Genetic Programming developed from Genetic Algorithms is one of the most well known approaches belonging to the latter. In this paper, we give detailed algorithm description as well as data construction that are necessary for learning single agent strategies at first. In following step moreover, we will extend developed methods into multiple robot domains. game. We investigate and contrast two different methods-simple team learning and sub-group loaming and conclude the paper with some experimental results.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2002.10a
• /
• pp.41-46
• /
• 2002

In the face of global competition, the requirements fer the continuously increasing productivity, flexibility and quality (dimensional accuracy, mechanical properties and surface properties) have imposed a major change on steel manufacturing industries. The automation of hot rolling process requires the developments of several mathematical models for simulation and quantitative description of the industrial operations involved. To achieve this objectives, a new loaming method with neural network to improve the accuracy of rolling force prediction in hot rolling mill is developed. Also, Genetic Algorithm(GA) is applied to select the optimal structure of the neural network and compared with that of engineers experience. It is shown from this research that both structure selection methods can lead to similar results.