Figures & data
FIGURE 1 Two activation functions: (a) sigmoid ; (b)
.
FIGURE 2 The backpropagation neural network with one hidden layer.
FIGURE 3 Gaussian for three differents variances.
FIGURE 4 Cascade correlation network with two hidden layers. The symbols denotes a neuron.
FIGURE 5 Sample of test-functions for training: (a) non-similar functions; (b) similar functions.
TABLE I Training results for the neural networks used for initial condition reconstruction
TABLE II Activation results for the noiseless experimental data
FIGURE 6 Reconstruction using multilayer perceptron NN with noiseless data.
FIGURE 7 Reconstruction using radial basis function NN with noiseless data.
FIGURE 8 Reconstruction using cascade correlation NN with noiseless data.
FIGURE 9 Reconstruction using multilayer perceptron NN with 5% of noise.
FIGURE 10 Reconstruction using radial basis function NN with 5% of noise.
FIGURE 11 Reconstruction using cascade correlation NN with 5% of noise.
TABLE III Activation results for the experimental data with 5% of noise
Table
