Browsing by Author "Shiue, YR"

Now showing 1 - 5 of 5

A power-spectrum based neural fuzzy connection admission mechanism for ATM networks
(1997)
Chang, CJ
;
Lin, SY
;
Shiue, YR
;
Cheng, RG
;
交大名義發表
ATM (asynchronous transfer mode) networks support services with bursty traffic. A sophisticated and real-time admission controller is needed not only to guarantee the required quality-of-service (QoS) for existing calls but also to achieve higher system efficiency. In this paper, we propose a power-spectrum based neural fuzzy connection admission control (PNFCAC) mechanism for an ATM network. Power-spectrum of input process explicitly describes the correlation behavior of input traffic and has a great impact on the queueing behavior of an ATM net work. The neural fuzzy controller has been successfully applied to deal with the complex problem which can not be analyzed by classical methods based on probability theory and bivalent logic for recent years. Under the constraint of QoS, we construct a decision hyperplane of the connection admission control according to parameters of the power spectrum. We further adopt the learning/adapting capabilities of the neural network to adjust the optimum location of the boundary between these two decision space. Simulation results show that the PNFCAC mechanism provides a superior system utilization over the conventional CAC schemes and the neural fuzzy CAC using time-domain analysis, while keeping the QoS contract.
An enhanced knowledge representation for decision-tree based learning adaptive scheduling
(2003-01-01)
Shiue, YR
;
Su, CT
;
工業工程與管理學系
;
Department of Industrial Engineering and Management
The classical decision tree (DT) learning approach to constructing DT knowledge bases is usually not considered if there exist some irrelevant and redundant attributes in the problem domain. Since the essential attributes are uncertain in manufacturing systems, how to select important manufacturing attributes to improve the generalization ability of knowledge bases and avoid overfitting training data in DT-based learning is a crucial research issue for the adaptive scheduling problem domain. In this study, we will first develop an attribute selection algorithm based on the weights of artificial neural networks (ANNs) to identify the importance of system attributes. Next, we will use the C4.5 DT learning algorithm to learn the whole set of training examples with important attributes in order to enhance knowledge representation. This hybrid ANN/DT approach is called an attribute selection DT (ASDT) based learning adaptive scheduling system. The results from the case study show that the use of an attribute selection algorithm to build scheduling knowledge bases delivers better generalization ability than in the absence of the attribute selection procedure in terms of the size of DTs under various performance criteria. Consistent conclusions are drawn from the resulting prediction accuracy of unseen data. The resulting prediction accuracy of unseen data also reveals that scheduling knowledge bases by the proposed attribute selection approach to constructing DTs can avoid overfitting the training data compared with the classical DT learning approach.
Attribute selection for neural network-based adaptive scheduling systems in flexible manufacturing systems
(2002)
Shiue, YR
;
Su, CT
;
工業工程與管理學系
;
Department of Industrial Engineering and Management
System attribute selection is an integral part of adaptive scheduling systems. Owing to the existence of irrelevant and redundant attributes in manufacturing systems, by selecting the important attributes, better performance or accuracy in prediction can be expected in scheduling knowledge bases. In this study, we first propose an attribute selection algorithm based on the weights of neural networks to measure the importance Of system attributes in a neural network-based adaptive scheduling (NNAS) system. Next, the NNAS system is combined with the attribute selection algorithm to build scheduling knowledge bases. This hybrid approach is called an attribute selection neural network-based adaptive scheduling (ASNNAS) system. The experimental results show that the proposed ASNNAS system works very well, when measured by a variety of performance criteria, as opposed to the traditional NNAS system and a single dispatching strategy. Furthermore, the scheduling knowledge bases in the ASNNAS system can provide a stronger generalisation ability compared with NNAS systems under various performance criteria.
Intelligent scheduling controller for shop floor control systems: a hybrid genetic algorithm/decision tree learning approach
(2003-08-01)
Su, CT
;
Shiue, YR
;
工業工程與管理學系
;
Department of Industrial Engineering and Management
This work develops an intelligent scheduling controller (ISC) to support a shop floor control system (SFCS) to make real-time decisions, robust to various production requirements. Selecting near-optimal subset system attributes (or features) based on various production requirements to construct ISC knowledge bases is a critical issue because of the existence of much shop floor information in an SFCS. Accordingly, this work developed a learning-based ISC methodology to acquire knowledge of a dynamic dispatching rule control mechanism. The proposed approach integrates genetic algorithms (GAs) and decision trees (DTs) learning to evolve a combinatorial optimal subset of features from possible shop floor information concerning a DT-based ISC knowledge classifier. A GA is employed to search the space of all possible subsets of a large set of candidate features. For a given feature subset, a DT algorithm is invoked to generate a DT. Applying the GA/DT-based knowledge learning mechanism to the experimental results demonstrates that the use of an optimal subset of system attributes to build scheduling knowledge bases enhanced generalization ability of the learning bias above that in the absence of an attribute selection procedure, in terms of prediction accuracy of unseen data under various performance criteria. Furthermore, simulation results indicate that the GA/DT-based ISC improves system performance in the long run over that obtained with classical DT-based ISC and the heuristic individual dispatching rule, according to various performance criteria.
PSD-based neural-net connection admission control
(1997)
Chang, CJ
;
Lin, SY
;
Cheng, RG
;
Shiue, YR
;
交大名義發表
ATM (asynchronous transfer mode) systems can support services with bursty traffic. An ATM system needs a sophisticated and real-time connection admission controller not only to guarantee the required quality-of-service (QoS) for existing calls but also to raise the system efficiency. Input process has a power-spectral-density (PSD) which explicitly contains the correlation behavior of input traffic and has a great impact on the system performance. Also, a neural network has been widely applied to deal with traffic control related problems in ATM systems because of its self-learning capability. In this paper, we propose a PSD-based neural-net connection admission control (PNCAC) method for an ATM system. Under the QoS constraint, we construct a decision hyperplane of the connection admission control according to parameters of the power spectrum. We further adopt the learning/adapting capabilities of the neural network to adjust the optimum location of the boundary between these two decision spaces. Simulation results show that the PNCAC method provides a superior system utilization over the conventional CAC schemes by an amount of 18%, while keeping the QoS contract.