-
Notifications
You must be signed in to change notification settings - Fork 5
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Allow optimization based acquisition functions #6
Comments
Hi @kiudee, |
Hi @michaelosthege, thank you for the reference. From having a cursory glance at the paper, it looks like it should be straightforward to implement. |
The paper was accompanied by a MATLAB implementation for which I also found a Python port. Unfortunately, due to the code formatting & absence of useful docstrings/comments I found it hard to understand. I have trouble understanding which values I'll have to pass & which constraints this approximation may put on my GP model.. |
In your case I would look at quadrature fourier features to implement the Thompson sampling part: Use it in conjunction with predictive variance reduction search (PVRS): edit: In bayes-skopt PVRS is implemented as follows: bayes-skopt/bask/acquisition.py Lines 278 to 308 in d7fa3d2
Note, however that this is sampling from a random set of points instead of using fourier features. |
hi @kiudee, My impression is that the QFF make some assumptions about independence of the problem dimensions, that we don't feel comfortable about with our problem. (That might change though, if we run into problems with computational complexity 😅)
|
Computation of acquisition functions only on sampled points is problematic in high-dimensional spaces, where the distance to the true optimum (of the acquisition function) will be large on average.
The text was updated successfully, but these errors were encountered: