Add theory page on the Bayesian formulation and contrast with SNAPHU - #75
Open
scottstanie wants to merge 2 commits into
Open
Add theory page on the Bayesian formulation and contrast with SNAPHU#75scottstanie wants to merge 2 commits into
scottstanie wants to merge 2 commits into
Conversation
Explains why the Carballo cost model makes whirlwind a Bayesian unwrapper: the posterior probability over the integer correction field that the min-cost flow maximizes, the hierarchical Lee/slope/coherence model with nuisance marginalization behind the p0/p1 tables, and how this differs from SNAPHU's plug-in density costs (genuinely ML in the smooth mode NISAR production runs). Includes the interferogram regimes where the two formulations should diverge, plus honest caveats about the clamp, unit-capacity linearization, and independence approximation. Linked from the nav, index, ALGORITHM.md, and ATBD section 2.3. https://claude.ai/code/session_018TRLW5tmLFsaJY55Apsauw
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Explains why the Carballo cost model makes whirlwind a Bayesian
unwrapper: the posterior probability over the integer correction field
that the min-cost flow maximizes, the hierarchical Lee/slope/coherence
model with nuisance marginalization behind the p0/p1 tables, and how
this differs from SNAPHU's plug-in density costs (genuinely ML in the
smooth mode NISAR production runs). Includes the interferogram regimes
where the two formulations should diverge, plus honest caveats about
the clamp, unit-capacity linearization, and independence approximation.
Linked from the nav, index, ALGORITHM.md, and ATBD section 2.3.
https://claude.ai/code/session_018TRLW5tmLFsaJY55Apsauw