Why we need so many neurons?
About neurons in MLP
hey @ishaan_123,
in deep learning we need to do any kind of feature engineering on our own manually as we want our model to do this task for us.
So to get it done by the model itself , it need number of nodes/neurons to understand the data and generate features that would be useful to get better results.
This is the reason we have so many neurons in a multi layer perceptron or any other deep learning model.
Note : It might be thought that then why don’t we always a very large number of neurons to get better results , the reason is sometimes when using crores of neurons in a simple task , it learns the data so well that if we bring a new unseen data to it , it gets confused and cant understand what to do and hence performs really bad. So we need to find the right number of neurons within a range to get the best results.
I hope this would have cleared your doubt.
Thank You and Happy Learning. 
I mean to say if we initailize our first input matrix w1 with all zeros then in first hidden layers all neurons have same value because every input is connected to each neuron
yes , but along with this you also need to have constant weights and bias to get same values in first hidden layer , which doesn’t happen generally.
I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.
On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.