I am wondering how to make a good neural network architect ?
how to choose that how many layers and units is needed to make a neural network?
Is there any way to do that?
The method is experimentation, and it’s a skill you develop with experience.
1 Like
so basically you are trying different amount of layers and units within, watching the changes in the loss function, am I right?
is there any ‘protocol’ how you or other specialists and professionals are performing these experiments on layers/unit amounts? can you maybe suggest any reading on that topic?
Thank you!
Yes, trying different layers and unit and measuring Loss on train and cv data help to
choose good neural network for deep learning ,Since there is no specific method of choosing a neural network that predicts best result .