Neural Networks

神级网络算法：

1. Multilayer Feed-Forward Neural Network

Including input layer, hidden layer, output layers

Every layers is informed by unites.

Input layers is inputted by training data set.

 2. Designing

1） confirm the numbers of layers

2）normalize input values between 1 and 0;

3) If the input values are discrete type, we can make every input as the characteristic value.

For instance, there is red, blue and yellow, three colors.

If red is characteristic value, so a0 is regarded as red and a0=1, a1 and a2 are both 0;

it can solve classification and regression.