Repository navigation
Hand each channel pair's factor-width-2 V to approx-chol 0.7 - #482
Merged
Merged
Conversation
Collaborator
Author
|
@s3alfisc @leostimpfle @johannes-schmieder @leninworld, this moves a lot of the logic from within to approx-chol. This fixes a wrong placed cut (see #481), removes more than 1000 production code lines, simplify within a lot, but will also introduce major conflicts on your forks. I am sorry for the inconvenience:( |
Each channel pair's Gram matrix is V Vᵀ, one column per observation with its two levels' loadings scaled by √w. approx-chol 0.7 takes that bipartite factor-width-2 input and does the SGDD → SDDM reduction itself, so each of its irreducible blocks becomes a Schwarz subdomain behind BlockSolver. ApproxSchurConfig::seed is removed: the sampler draws from ApproxCholConfig::seed. The v21 wire fixture is regenerated. Refs #481
Block elimination, the cross tables, the SDDM conversion with its scaling and grounding, the Schur assembly and the CSR block are unreferenced since the reduction moved into approx-chol. Refs #481
schroedk
force-pushed
the
feat/approx-chol-0.7
branch
from
October 31, 2026 15:23
2b35859 to
e6bd37e
Compare
…oundary approx-chol 0.7 reports any deficit past rounding, so filter on violation > tolerance as before. Document dense_threshold as the Cholesky dimension with a signed block's cover sheets counted.
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #481, Closes #454
Each channel pair passes approx-chol 0.7 its Gram factor
V: one column per observation, with both levels' loadings scaled by√w. approx-chol does the SGDD→SDDM reduction itself, and each irreducible block it returns becomes a Schwarz subdomain.domain/factor_pairs.rsbuildsVfrom level, loading and weight columns, and keeps pair enumeration, isolated-level batching and the partition of unity.block_solver.rsadapts approx-chol'sBlockFactors toLocalSolver.block_elim/,csr_block.rs,domain/cross_tab/,factor_pairs/{grounding,sddm}.Breaking:
ApproxSchurConfig::seed(Pythonseed=) is removed; the sampler draws fromApproxCholConfig::seed.ScalingConfig::tolerancemust lie in[0, 1).Notes:
Tests: all 21 within test binaries pass, including bitwise determinism across thread counts. pytest: 116 passed. No benchmark has been run yet.