How does JMP initiate the values of w and b for logistic regression loss function? How to simulate the same iterations as JMP reports?

[#1]

I have got some questions about the LR model.

baiyun_0-1628735743565.png

- How does JMP initiate the values of w and b for logistic regression loss function? 

- How does Relative Gradient derive(formula of this value)?

- What is the difference between Relative Gradient and NormGradient?

- Which parameter does this gradient value represent? w or b?

- Why no mention of the gradient update of the b(interception) parameter? 

I just want to simulate the iteration as JMP did, the above questions stuck me from repeating this iteration. Look forward to your awesome reply!

[#2]

Hi,

I've searched the JMP Help documentation but I can't find much explanation of the Iterations report:

https://www.jmp.com/support/help/en/16.0/#page/jmp/iterations.shtml#

I suggest that you contact JMP Technical Support to find out more.

Phil