Feat: Dynamic learning rate decay

[maxdinkel]

Description and Context

This PR adds a step-wise learning rate decay for Stochastic variational inference and the dynamic learning rate decay method from my upcoming paper.

Related Issues and Merge Requests

• Closes
• Blocks
• Is blocked by
• Follows
• Precedes
• Related to
• Part of
• Composed of

How Has This Been Tested?

Checklist

• My commit messages mention the appropriate issue numbers (advised but optional).
• I mention the appropriate issue numbers in this MR.
• I updated documentation where necessary.
• I updated the README where necessary
• I have added as (few,) small and fast tests as possible to cover my changes.

Additional Information

Interested Parties

Possible reviewers:

Other interested parties:

[gilrrei]

Looks good to me 👍