Good afternoon
I want to ask a question
Does the shape of the data distribution affect the convergence of the model ?
For example
If I have the data distribution bellow and I want to fit a circle shaped model that surround the blue dots , does this distribution’s shape have any relation with the model being converging slowly
Yes, definitely shape affects the convergence
Shape is direct relation from degree of polynomial and as we know degree of polynomial definitely affects convergence
In this situation, the time for convergence has a lot to do with whether you normalized the features, and what learning rate you select.