Chunkai Zhang

Publications (6)0 Total impact

• Chapter: An Evolved Recurrent Neural Network and Its Application

  Chunkai Zhang, Hong Hu
  11/2006: pages 265-283;
• Conference Proceeding: An evolved recurrent neural network and its application in the state estimation of the CSTR system

  Chunkai Zhang, Hong Hu
  [show abstract] [hide abstract]
  ABSTRACT: Continuous stirred tank reactor system (CSTR) is a typical chemical reactor system with a complex nonlinear dynamic characteristics. In this paper, a recurrent neural network (RNN) evolved by a cooperative scheme is proposed to estimate the state of the CSTR system, which combines the architectural evolution with weight learning. In this scheme, particle swarm optimization (PSO) adoptively constructs the network architectures, then evolutionary algorithm (EA) is employed to evolve the network nodes with this architecture, and this process is automatically alternated. It can effectively alleviate the noisy fitness evaluation problem and the moving target problem. In addition of these, a closer behavioral link between the parents and their offspring is maintained, which improves the efficiency of evolving RNN. The results show that the proposed scheme is able to evolve both the architecture and weights of RNN, and the effectiveness and efficiency is better than the algorithms of TDRB, GA, PSO, and HGAPSO applied to the fully connected RNN.
  Systems, Man and Cybernetics, 2005 IEEE International Conference on; 11/2005
• Conference Proceeding: Using PSO algorithm to evolve an optimum input subset for a SVM in time series forecasting

  Chunkai Zhang, Hong Hu
  [show abstract] [hide abstract]
  ABSTRACT: Using particle swarm optimization (PSO) algorithm to evolve an optimum input subset for a SVM is proposed Binary PSO algorithm is employed in feature selection, in which each particle represented as a binary vector corresponds to a candidate input subset. A swarm of particles flies through the input set space for targeting the optimal subset. In order to evaluate the reasonable fitness of each input subset, PSO algorithm is used to adoptively evolve SVM to obtain the best performance of network, in which each particle represented as a real vector corresponds to the candidate kernel parameters of SVM. This method has been applied in a real financial time series forecasting, the results show that it has better performance of generalization, and higher rate of convergence.
  Systems, Man and Cybernetics, 2005 IEEE International Conference on; 11/2005
• Article: An ANN's Evolved by a New Evolutionary System and Its Application

  Chunkai Zhang, Huihe Shao
  [show abstract] [hide abstract]
  ABSTRACT: The paper presents a new evolutionary system for evolving artificial neural networks (ANN's). In the process of evolution, the network architecture and the node weights of ANN's are evolved alternately, and the evolution value of network architecture is related to the error value of ANN's evolved by node weights. An evolved ANN's has been used in modelling product quality estimator for a fractionator of the hydrocracking unit in the oil refining industry, the results show that it has good accuracy and generalisation ability. 1 Introduction Design of a near optimal ANN's architecture can be formulated as a search problem in the architecture space where each point represents a type of architecture. Miller et al. indicated that the evolutionary algorithms are better candidates for searching architecture space than those constructive and pruning algorithms [1, 2]. GA was used to evolve ANN's, but the evolution of ANN's often suffers from the permutation problem [3]. This problem...
  12/2000;
• Conference Proceeding: Particle swarm optimisation for evolving artificial neural network

  Chunkai Zhang, Huihe Shao, Yu Li
  [show abstract] [hide abstract]
  ABSTRACT: The information processing capability of artificial neural networks (ANNs) is closely related to its architecture and weights. The paper describes a new evolutionary system for evolving artificial feedforward neural networks, which is based on the particle swarm optimisation (PSO) algorithm. Both the architecture and the weights of ANNs are adaptively adjusted according to the quality of the neural network. This process is repeated until the best ANN is accepted or the maximum number of generations has been reached. A strategy of evolving added nodes and a partial training algorithm are used to maintain a close behavioural link between the parents and their offspring. This system has been tested on two real problems in the medical domain. The results show that ANNs evolved by PSONN have good accuracy and generalisation ability
  Systems, Man, and Cybernetics, 2000 IEEE International Conference on; 02/2000
• Chapter: Feature Selection in SVM Based on the Hybrid of Enhanced Genetic Algorithm and Mutual Information

  Chunkai Zhang, Hong Hu
  [show abstract] [hide abstract]
  ABSTRACT: Feature selection is a well-researched problem, which can improve the network performance and speed up the training of the network. In this paper, we proposed an effective feature selection scheme for SVM using the hybrid of enhanced genetic algorithm and mutual information, in which mutual information between each input and each output of the data set is employed in mutation in evolutionary process to purposefully guide search direction based on some criterions. In order to avoid the noise fitness evaluation, in evaluating the fitness of an input subset, a SVM should adaptively adjust its parameters to obtain the best performance of network, so an enhanced GA is used to simultaneously evolve the input features and the parameters of SVM. By examining two real financial time series, the simulation of three different methods of feature selection shows that the feature selection using the hybrid of GA and MI can reduce the dimensionality of inputs, speed up the training of the network and get better performance.
  01/1970: pages 307-316;