Fisher Information for Instruments #503
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
added 2 commits
October 22, 2024 23:30
Updates the implementation of the fisher information calculation to support instruments. In addition to this there is a rework to the way that models are regularized to remove the use of depolarization and instead add in a minimum probability used for clipping small values. This was done in part to avoid the need to add in extra handling for regularizing instruments. Seems to still give sensible results for models with known fisher info spectra. Also patches back in support for the 'cumulative' kwarg when computing fisher information by L (I could've sworn we had this implemented, but maybe it got lost at some point?).
Add unit tests for instrument fisher information and modify existing unit tests. It looks like the change in the regularization likely adds some additional numerical gunk to the accumulation process which was resulting in small, but nonzero differences in the FIM matrices depending on if they were accumulated all at once, or length by length. Or at least that is my best guess for what is going on, hopefully that isn't incorrect.
sserita
approved these changes
Nov 5, 2024
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Outdated suggestions cannot be applied.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.
TLDR: This PR updates the implementation of the fisher information for experiment designs to support models/circuits with quantum instruments. This was requested ages ago in #464.
As part of updating the fisher information related code to support instruments this PR also changes the way that we do regularization. Previously this regularization was managed by adding a small amount of depolarizing error to the SPAM operations. Following a recommendation from @robinbk this has been changed to instead perform regularization by setting a minimum value for the probability p_min such that any probability below this threshold is clipped to this minimum value. As such, while in my testing the results looked nearly identical to previously calculated fisher information matrices for known experiment designs, the new code won't exactly reproduce the results of the older implementation.