How Backpropagation Works? (Cont.)

Step 3. Calculating the Updated Weight Value
Now, let’s put all the values together:


Let’s calculate the updated value of W5 (optionally multiplied by some learning rate, η, which we’ll set to 0.5):

  1. Similarly, we can calculate the other weight values as well.
  2. After that we will again propagate forward and calculate the output. Again, we will calculate the error.
  3. If the error is minimum we will stop right there, else we will again propagate backwards and update the weight values.
  4. This process will keep on repeating until error becomes minimum.
Conclusion
The pseudocode for the backpropagation is as follows:

Backpropagation Algorithm 
initialize network weights (often small random values)
do {
  foreach training example named ex {
    // forward pass
    prediction = neural-net-output( network, ex )
    actual = teacher-output( ex )
    compute error( prediction - actual ) at the output units
    // backward pass
    for all weights from hidden layer to output layer
      compute { displaystyle Delta w_{h} } 
    // backward pass continued
    for all weights from input layer to hidden layer
      compute { displaystyle Delta w_{i} }
    // input layer not modified by error estimate
    update network weights 
  }
}
until all examples classified correctly or another stopping criterion satisfied
return the network




      Q: How do you make a skunk stop smelling?    
      A: Pinch its nose closed.