[mlpack] Draft proposal on Enhancing CMA-ES

[Ryan Curtin]

On Sun, Mar 29, 2020 at 11:38:08AM +0800, Yihan Wang wrote:
> Dear mentors,
> 
> I have just submitted a draft proposal, and I will appreciate it if you
> have to time to comment on it.
> 
> And I also have a problem regrading to the project. In ACM-ES, we need to
> add a surrogate model, which in the original paper is SVM. I have found
> that there is an SVM implementation in mlpack, but I am not sure if it is
> ok to include a model from the mlpack repo. If there is any suggestion on
> this problem?

Hey Yihan,

Thanks for the submission.  I'll see if I have time to review it, but do
remember, mentors can ask questions after the proposal deadline, so it
doesn't need to be 100% perfect on the day of submission. :)

For SVM, I agree, we should avoid including mlpack code in ensmallen (it
would make a circular dependency).  Maybe it's easier to just implement
the SVM objective function in ensmallen_bits/problems/ for testing?  We
did the same thing for the logistic regression objective function to
avoid the circular dependency.

Hope this helps.

Thanks!

Ryan

-- 
Ryan Curtin    | "Are those... live rounds?"
ryan at ratml.org | "Seven-six-two millimeter.  Full metal jacket."