Batch capable sampling functions + proto-type HMC/MCMC #25
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
|
Partially converged chain example, with truth highlighted.
|
|
OOOOooooohh Sweeet! |
|
If someone has a chance, I could use a second set of eyes on the loss function to figure out why it seems to require me to multiply it by a tiny number to get the chain to move :-/ I assume there is a bug somewhere... |
|
It appears some of the issues go away when I reduce the MCMC step-size or the HMC initial step-size... |
Merge branch 'batch_hmc' of https://github.com/DifferentiableUniverseInitiative/DHOD into batch_hmc
|
I'm gonna flag this as out of date most likely, and will close this PR |
|
bits of codes were already merged from the ben_2021 branch |
|
Ah, I think we just need the notebook with the standard MCMC |
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Outdated suggestions cannot be applied.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.
I spent a few hours going through the various sampling related functions to make them take arrays of values. In order to get the overall likelihood working I also had to change around the power-spectra code. Right now I use map_fn to "broadcast" (appy?) two the functions over the batch (the bin-count and fft functions). Not sure if this is needed though.
I also include some MCMC and HMC examples; currently things don't look too great and it is still not particularly fast. A few things that limit speed are...
Also, my normalization of my likelihood is 100% incorrect; when I was trying what I thought was the correct thing my chains never moved :(