EKF (diagonal) #9
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
There was a problem hiding this comment.
we shouldn't include files like these in this PR. They seem unrelated to the PR. We should open a second PR for this, like "update sghmc with new library"
There was a problem hiding this comment.
Yeah this was lazy from me, the flexi_tree_map update should have been its own PR
johnathanchiu
approved these changes
Jan 31, 2024
There was a problem hiding this comment.
I like what you did with the flexi_tree_map. Cleans up the code nicely.
I'll trust @dmitrisaberi's review on the EKF stuff. I still need to read through the paper more closely to understand the methods.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adds extended Kalman filter (EKF) implementations with diagonal Hessian and empirical Fisher approximations to the Hessian of the likelihood.
It's perhaps a non-standard version of the EKF which uses the gradient and Hessian approximation of the likelihood at time t rather than the Jacobian of the non-linear conditional mean of the likelihood. This version is easier to implement since it only requires the log_likelihood function itself (and its gradients) rather than the conditional mean of the likelihood function. For more details see https://arxiv.org/abs/1703.00209 which also gives a gradient descent interpretation and an equivalence between likelihood inverse temperature and learning rate (which we adopt).
Additionally this PR adds a
flexi_tree_mapfunction to simplify theinplacecode.