A Modified Differential Evolution (MDE) was proposed, which based on the basic Differential Evolution (DE) algorithm principle and implementing framework of DE. Optimizing the initial individuals with the 1/2 rule, and then introducing the reorganization of Evolution Strategies during the period of mutation...

A novel neural network approach is presented for solving nonlinear bilevel programming problem. The proposed neural network is proved to be Lyapunov stable and capable of generating optimal solution to the nonlinear bilevel programming problem. The asymptotic properties of the neural network are analyzed...

Based on analysis of conventional back-propagation (BP) network, the causes of error curve oscillation and excessive learning are first proposed. Next, a new BP with the error contracting gradually algorithm is put forward, through setting up neuron error threshold function, only when neuron’s error...

Pulse Coupled Neural Network (PCNN), called the third generation of neural network, is widely used in image processing for its basic characteristics of coupling mechanism and achieved some results. PCNN is improved based on the above basic characteristics in the paper: correlation coefficient is used...

Neural network has a powerful parallel processing capability, along with its rise in the fuzzy risk cluster analysis which has occupied an important position, however, the quality of fuzzy risk clustering results is influenced by the initial value of options. The initial cluster center method for fuzzy...

Benchmarking of best practices has been proved usefully in the business and manufacturing sectors. However, benchmarking is not established in the life cycle cost management (LCCM) of building project. This paper firstly analysis the barriers of LCCM and finds that the important barriers are history...

In view of the disadvantages of the traditional time-varying algorithm about nonstationary random vibration signal of a spacecraft with close spaced modal frequency. A process neural network (PNN) based on the empirical mode decomposition (EMD) method is put forward. First, the EMD method is utilized...

This paper presents a new variable step size LMS algorithm based on neural network (BP-LMS). A non-linear relationship amongst the input vectors, deviation errors and the learning steps is constructed by BP model, which is employed to determine the learning steps during adaptive processing. The proposed...

In order to conquer the localization of the multi-layered feed forward neural networks, this paper presents a kind of neuron models utilizing information of local samples—the UILS neuron model, including an adaptive neuron model and a self-organization neuron model. Differing from traditional models,...

This paper proposes a new approach to improve the control precision of shaking table control system, in which the fuzzy neural network (FNN) technique and iterative learn control (ILC) are combined and developed a new control technique. A FNN inverse model is built and is identified through a white noise...

A nonlinear model predictive control (NMPC) algorithm based on a BP-ARX combination model is proposed for multivariable nonlinear systems whose static nonlinearity between inputs and outputs could be obtained. The dynamic behavior of the system is described by a parameter varying ARX model, whose parameters...

To evaluate the enterprises' strategic management performance scientifically and accurately, this paper proposes the improved BP neural network model which imports the adjustable activation function and the Levenberg-Marquardt optimization algorithm. The improved model not only can simulate the expert...

For the reason of different images with different space distribution of gray levels, we proposed a texture representation based on simplified pulse coupled neural networks (PCNN) model which output a series of binary images corresponding to different gray levels. Then we transformed the images into 1D...

A class of Cohen-Grossberg neural networks with distributed delays are considered. By using the coincidence degree theorem and differential inequality techniques, sufficient conditions for the existence and exponential stability of the periodic solutions are established, Without assuming the boundedness,...

The batch-to-batch model-based iterative optimal control strategy for batch processes is realized based on multilayer recurrent fuzzy neural network (MRFNN) and chaotic search. MRFNNs are used to model batch processes. Modeling and optimization problems are mainly solved by chaotic search. Due to model-plant...

Ignoring the samples far away from the training samples, our study team gives a new norm-based derivative process of localized generalization error boundary. Enlightened by the above research, this paper proposes a new method to construct radial basis function neural networks, which minimizes the sum...

We develop a novel synthesis way to effectively generate CMOS spiral inductor’s layout parameters using artificial neural network and genetic algorithm. An accurate neural network model for CMOS spiral inductors is firstly developed based on measured results from TSMC 0.13um MM/RF process with the frequency...

Based on the theory of topological degree and properties of M-matrix, by constructing proper vector Lyapunov functions, the existence and uniqueness of the equilibrium point and its global exponential stability are investigated for a class of neural networks with unbounded and varying delays. Without...