00:04Come one layer in using those we have found out the derivative of the loss function with
00:10respect to you know the weights that are present in the inside layers. So, it is a propagation
00:16or the error is propagated in the backward direction. Then that is the reason it is called
00:22as back propagation algorithm. So, with this one I come to the end of today's
00:26discussion. Just to summarize what we have discussed today, we started with a toy neural
00:32network with which we have set up for explaining the process of back propagation, how the partial
00:39derivatives of the loss function are calculated with respect to the weights of the neural network.
00:44In the toy network we have taken very simple activation function like sigmoid activation
00:49function and very simple loss functions like mean squared error. And we have calculated we
00:54have shown the calculation of gradient of loss function with respect to one of the weights
01:01or the partial derivative of loss function with one of the weights in the last but one
01:05layer. And then we have taken again the partial derivative of the loss function with respect
01:10to another weight which is there inside in the internal layers. So, with this understanding
01:17we will be able to do the partial derivative calculations of the loss function with respect to all the weights
01:24that are there in the neural network. The similar way the bias also gets updated and we have seen
01:31how the weight updation happens at the end of the gradient descent algorithm also. So, that is the end
01:41of today's lecture session. I hope you have understood the mathematical formulation of back propagation why
01:48it is called as back propagation where all we are using the back propagated errors in finding
01:55out the gradient of the loss function. So, with this one I come and conclude today's session. In the next
02:02session we will talk about variants of gradient descent algorithm and momentum based optimizers which
02:13are used in machine learning model. Till that time keep learning. Thank you all.
Comments