Deep Learning: Practical Neural Networks with Java

Implementing a hybrid neural network

Now, let's implement a simple code that can be used in the neuro-fuzzy and neuro-genetic networks. First, we need to define Gaussian functions for activation that will be the membership functions:

public class Gaussian implements ActivationFunction{
  double A=1.0,B=0.0,C=1.0;
  public Gaussian(double A){ ///…
  }
  public double calc(double x){
    return this.A*Math.exp(-Math.pow(x-this.B,2.0) / 2*Math.pow(this.C,2.0));
  }
}

The fuzzy sets and rules need to be represented in a way that a neural network can understand and drive the execution. This representation includes the quantity of sets per input, therefore having the information on how the neurons are connected; and the membership functions for each set. A simple way ...