The purpose of backpropagation is to update each of the weights in the network so that they cause the actual output to be closer to the target output, thereby minimizing the error for each output neuron and the network as a whole.
Let's focus on an output layer first. We are supposed to find out the impact of change in w5 on the total error.
This will be decided by
. It is the partial derivative of Etotal with respect to w5.
Let's apply the chain rule here:
= 0.690966 – 0.9 = -0.209034
= 0.213532
InputOL1 = w5*OutputHL1 + w7*OutputHL2 + B2
= 0.650219
Now, let's get back to the old equation:
To update the weight, we will use the following formula. We have set the learning rate to be α = 0.1:
Similarly,
are supposed to be calculated. The approach remains the same. We will leave this to compute as it will help you in understanding the concepts better.
When it comes down to the hidden layer and computing, the approach still remains the same. However, the formula will change a bit...