Showing papers by "Xiaoou Li published in 2007"

Passivity Analysis of Dynamic Neural Networks with Different Time-scales

Abstract: Dynamic neural networks with different time-scales include the aspects of fast and slow phenomenons. Some applications require that the equilibrium points of the designed networks are stable. In this paper, the passivity-based approach is used to derive stability conditions for dynamic neural networks with different time-scales. Several stability properties, such as passivity, asymptotic stability, input-to-state stability and bounded input bounded output stability, are guaranteed in certain senses. A numerical example is also given to demonstrate the effectiveness of the theoretical results.

SVM Classification for Large Data Sets by Considering Models of Classes Distribution

Abstract: Despite of good theoretic foundations and high classification accuracy of support vector machines (SVM), normal SVM is not suitable for classification of large data sets, because the training complexity of SVM is very high. This paper presents a novel SVM classification approach for large data sets by considering models of classes distribution (MCD). A first stage uses SVM classification in order to gets a sketch of classes distribution. Then the algorithm obtain the support vectors (SVs) most close between each class and construct a ball using minimum enclosing ball from each pair of SVs with different label. The data points included in the balls constitute the MCD, which is the framework in the boundary of each class and represents the most important data points, these data points are used as training data for a posterior SVM classification. Experimental results show that our approach has good classification accuracy while the training is significantly faster than other SVM classifiers.

Applying Petri Nets in Active Database Systems

Abstract: Reactive behavior of active database systems is achieved through the definition of event-condition-action (ECA) rules. Generally, ECA rule representation and processing are separated in the majority of existing active database systems. In this paper, we propose a conditional colored Petri net (CCPN) approach to model and simulate ECA rules. CCPN can not only integrate rule representation and processing in only one model, but is also independent of the actual database system. Furthermore, we have developed a software platform named ECAPNSim, which can generate a CCPN model automatically from a text file of ECA rule description, and communicate with a traditional database system when an event is detected from the database or an action command is generated by the CCPN simulator.

Two-stage svm classification for large data sets via randomly reducing and recovering training data

Abstract: Despite of good theoretic foundations and high classification accuracy of support vector machine (SVM), normal SVM is not suitable for classification of large data sets, because the training complexity of SVM is very high. This paper presents a novel two stages SVM classification approach for large data sets by randomly selecting training data. The first stage SVM classification gets a sketch of support vector distribution. Then the neighbors of these support vectors in original data set are used as training data for the second stage SVM classification. Experimental results demonstrate that our approach have good classification accuracy while the training is significantly faster than other SVM classifiers.

Recurrent fuzzy neural networks for nonlinear system identification

Abstract: In this paper, we propose a new recurrent fuzzy neural network, which has the standard state space form, we call it state-space recurrent neural networks. Input-to-state stability is applied to access robust training algorithms for system identification. Stable learning algorithms for the premise part and the consequence part of fuzzy rules are proved.

Multi-Class Support Vector Machines for Large Data Sets via Minimum Enclosing Ball Clustering

Abstract: Support vector machines (SVM) for binary classification have been developed in a broad field of applications. But normal SVM algorithms are not suitable for classification of large data sets because of high training complexity. This paper introduces a novel two-stage SVM classification approach for large data sets: minimum enclosing ball (MEB) clustering is introduced to select the training data from the original data set for the first stage SVM, and a de-clustering technique is then proposed to recover the training data for the second stage SVM. Then we extend binary SVM classification to case of multiclass. The novel two-stage multi-class SVM has distinctive advantages on dealing with huge data sets. Finally, we apply the proposed method on several benchmark problems, experimental results demonstrate that our approach have good classification accuracy while the training is significantly faster than other SVM classifiers.

Stable antiswing PD control for overhead crane systems with velocity estimation and uncertainties compensation

Abstract: A common problem in robot navigation is behaviour rule selection when more than one action of the same type is available. Here action selection or decision-making procedure incorporating α-level fuzzy logic is discussed and used for selecting an appropriate action during mobile robot navigation. In the present approach, the operational strategies of the human expert driver are transferred via fuzzy logic to the robot navigation in the form of optimising behaviour rules without decreasing the input space and resolved the conflicts similar to human thinking. This paper presents the mathematical aspects of resolving conflicts when more than one context rule of the same kind is in action. Simulation results are presented based on the real life situation. The techniques presented in this paper are validated by conducting real world experiments using Khepera miniature Robot.

Recurrent Fuzzy CMAC for Nonlinear System Modeling

Abstract: Normal fuzzy CMAC neural network performs well because of its fast learning speed and local generalization capability for approximating nonlinear functions. However, it requires huge memory and the dimension increases exponentially with the number of inputs. In this paper, we use recurrent technique to overcome these problems and propose a new CMAC neural network, named recurrent fuzzy CMAC (RFCMAC). Since the structure of RFCMAC is more complex, normal training methods are difficult to be applied. A new simple algorithm with a time-varying learning rate is proposed to assure the learning algorithm is stable.

Neural identification based on sliding mode observer

Abstract: In this paper, a new on-line neural identification method is presented. The identified nonlinear systems are partial-state measurement. Their inner states, parameters and structures are unknown. The design is based on the combination of a sliding mode observer and a neuro identifier. First, a sliding mode observer, which does not need any information of the nonlinear system, is applied to get the full states. Then a dynamic multilayer neural network is used to identify the whole nonlinear system. The main contributions of this paper are: (1) a new observer based identification algorithm is proposed; (2) a stable learning algorithm for the neuro identifier is given.

Verification of active rule base via conditional colored Petri nets

Abstract: Active rules are widely used in modern reactive software systems, such as active data base management systems, smart homes, etc. Determining if an active rule base is free of errors is an important process for both rule base design and maintenance. In this paper, we originally define the basic errors in an active rule-based system by extending the conceptions which are used generally in production rule base. Furthermore, a Petri net-based approach is proposed for active rule-base verification. An example on smart homes design is used as an application.

Detecting RNA sequences using two-stage SVM classifier

Abstract: RNA sequences detection is time-consuming because of its huge data set size. Although SVM has been proved to be useful, normal SVM is not suitable for classification of large data sets because of its high training complexity. A two-stage SVM classification approach is introduced for fast classifying large data sets. Experimental results on several RNA sequences detection demonstrate that the proposed approach is promising for such applications.

Integrated Analytic Framework for Neural Network Construction

Abstract: This paper investigates the construction of a wide class of singlehidden layer neural networks (SLNNs) with or without tunable parameters in the hidden nodes. It is a challenging problem if both the parameter training and determination of network size are considered simultaneously. Two alternative network construction methods are considered in this paper. Firstly, the discrete construction of SLNNs is introduced. The main objective is to select a subset of hidden nodes from a pool of candidates with parameters fixed `a priori'. This is called discrete construction since there are no parameters in the hidden nodes that need to be trained. The second approach is called continuous construction as all the adjustable network parameters are trained on the whole parameter space along the network construction process. In the second approach, there is no need to generate a pool of candidates, and the network grows one by one with the adjustable parameters optimized. The main contribution of this paper is to show that the network construction can be done using the above two alternative approaches, and these two approaches can be integrated within a unified analytic framework, leading to potentially significantly improved model performance and/or computational efficiency.

Fuzzy Modeling Via On-Line Clustering and Support Vector Machine

Abstract: This paper describes a novel fuzzy rule-based modeling approach for some slow industrial processses. Structure identification is realized by clustering and support vector machines. When the process is slow, fuzzy rules can be obtained automatically. Parameters identification uses the techniques of fuzzy neural networks. A time-varying learning rate assures stability of the modeling error.