Unsupervised learning: Training Procedure

• For each input:
  • Determine the winning output node
  • For each output node i within the radius of the winner
    • For each input node j
      • weight_ij = weight_ij + (learningRate * (input_ij - weight_ij))
• Determining the distance weight
  • Calculate the map distance to the winning node
  • Divide it by the radius
  • Subtract it from one; minimum value is zero, maximum value is one
• Determining the learning rate
  • Maintain the total distance weights that have been applied to the node
  • Learning rate is 1.0/total