 Wysłany: Pią 14:40, 24 Gru 2010 Temat postu: Particle Swarm Optimization Algorithm for the whol |
|
| Particle Swarm Optimization Algorithm for All
Network W = 1w,uggs bestellen,. , Wx. ~, Wx, Wx,ugg italia, 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,belstaff uomo, 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,microsoft outlook, will attempt to work with specific practical problems, and other classical test functions,Ed Hardy outlet, expanding across the connection-based multi-layer feedforward neural networks in various fields. |
|
