I did not understand how can we choose what will be the best values of w and b in a neural network and why exactly are they used.

It’s possible that I’m just misinterpreting your question, but my response is that the point is we do not *choose* the w and b values: the network *learns* the best values by using the combination of Back Propagation and Gradient Descent.

Thank you I understood now why we choose random values for w and b initially.

1 Like