// absolutely original object-oriented Artificial Neural Network Processing implementation (c) 2009 by Pablo Gindel

class Nnode {
  private float input[]; 
  float weight[];
  float output;
  Nnode (int num_inputs) {  
    input = new float [num_inputs]; 
    weight = new float [num_inputs]; 
  } 
  void set_inputs (float layer_input[]) {
    arrayCopy (layer_input, input);
  }  
  void set_weights (float layer_weight[]) {
    arrayCopy (layer_weight, weight);
  }
  void update () { 
    float accumulator = 0;
    for (int index=0; index<input.length; index++) {
      accumulator += input[index]*weight[index];
    }
    output = 1 / (1 + exp(-accumulator));                 //sigmoid    
  } 
} 
  
class Nlayer {
  Nnode node[]; 
  float input[];
  Nlayer (int num_nodes, int num_inputs) {  
    node = new Nnode[num_nodes];
    for (int index=0; index<node.length; index++) {
      node[index] = new Nnode(num_inputs);
    }
    input = new float [num_inputs];
  }
  void set_inputs (float layer_input[]) {
    arrayCopy (layer_input, input);
  }
  void update () { 
    for (int index=0; index<node.length; index++) {
      node[index].set_inputs(input);
      node[index].update();
    }
  } 
} 
  
class Nnetwork {
  Nlayer hidden_layer;
  Nlayer output_layer;
  float input[];
  Nnetwork (int output_layer_num_nodes, int hidden_layer_num_nodes, int num_inputs) {  
    hidden_layer = new Nlayer(hidden_layer_num_nodes, num_inputs);
    output_layer = new Nlayer(output_layer_num_nodes, hidden_layer_num_nodes);
    input = new float [num_inputs];
  }
  void update () { 
    hidden_layer.set_inputs(input);
    hidden_layer.update();
    for (int index=0; index<hidden_layer.node.length; index++) {
      output_layer.input[index] = hidden_layer.node[index].output;
    }
    output_layer.update();
  }
}

// instanciar una network: Nnetwork network (n_outs, n_hidden_nodes, n_inputs)
// los outputs de toda la network son: network.output_layer.node[].output
// los inputs de toda la network son: network.input[]
// los weights de toda la network son: network.xxx_layer.node[].weight[]
// para calcular el resultado de toda la network: network.update()