Particle Swarm Optimization Algorithm for the whol

e707004304

Particle Swarm Optimization Algorithm for All

Network W = 1w,[link widoczny dla zalogowanych],. , Wx. ~, Wx, Wx,[link widoczny dla zalogowanych], 1,}: f13.1248, -28.7716,13.1246, -28.4645,55.0249} (17) the traditional multi-layer feedforward neural network requires at least two hidden layer neural element to solve the XOR problem. Let the training accuracy of 1e-10, hidden layer and output layer activation function is a conventional s-type activation function, the performance comparison of two network shown in Table 1. Table 1 XOR of two networks for solving the problem of the performance comparison shown in Table 1, although based across the connection layer feedforward neural network using only a hidden element,[link widoczny dla zalogowanych], but the convergence is better than traditional networks, and has better stability . 5 Conclusion across the connection based on multilayer feedforward neural network, through the improvement of connection and discard all based on non-fully connected neural network model of traditional, helps reduce the complexity of the network structure, optimization of neural network structure a new way of thinking. Simulation results show that the convergence and stability of the network than the traditional network has improved greatly. Of course, this paper only the classical XOR problem shows the performance improvement in a further study,[link widoczny dla zalogowanych], will attempt to work with specific practical problems, and other classical test functions,[link widoczny dla zalogowanych], expanding across the connection-based multi-layer feedforward neural networks in various fields.