A more universal entropy calculation method for sampling based inference #151
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
…culation method by comparing estimates with an analytical computation.
ThijsvdLaar
approved these changes
Feb 18, 2021
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Outdated suggestions cannot be applied.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.
In the previous implementation of sampling based inference, the free energy calculation was not possible in smoothing tasks unless the hidden states are Gaussian distributed. With this PR, we generalise the entropy calculation to broader range of inference tasks such as smoothing in switching state space models where the forward messages, commute through equality nodes on the most upper chain, are constituted by SampleList. The entropy calculation is carried out by conveying the proposal distribution and integrand information on SampleList messages. When a SampleList message collides with a probability distribution or its pdf, our method updates the integrand. This allows entropy calculation not only for those cases where two probability distribution or pfd messages collide, i.e. m1(x).m2(x), but also for the cases when multiple collisions are required to approximate posteriors, e.g. q(x) \propto m1(x).m2(x).m3(x).m4(x).