Skip to content
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

Smtwcompkernel #653

Open
wants to merge 6 commits into
base: master
Choose a base branch
from

Conversation

NicolasJeanGonel
Copy link
Contributor

Added ways to combine kernels to create more complex kernels. For now only addition and multiplication are possible. Several important point should be noted :
As they are, Pow_Exp kernels are not usable in a combination as they lack the computation of the power for the distance and should be rewritten to include them.
In order to normalize the resulting kernels in the case of addition, a dummy kernel is multiplied to the resulting kernel which can be a bit heavy and confusing for the user.
Chained derivatives for the addition and multiplication are theoretically working but based on personnal tests felt very slow and inefficient, with disappointing results. Either there is a better way to compute them, or it is preferable to stay with gradient-free optimizer with a combined kernel.
It seems that these complex kernels can cause more local minimums, making it so that the order of addition and multiplication can create different results ; this issue is still investigated.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants