Conversation
fishidaho
marked this pull request as draft
September 18, 2026 03:36
`to_scipy_sparse()` (#47) returns the uncentered `Delta` term and leaves `means` for the caller to subtract. For the two recipes that do not center that is the normalized matrix; for the three that do (`parafac2`, `scanpy`, `pearson`) it is not, and the old docstring mentioned `means` only in passing, after describing the return value in terms that read as complete. Measured on a 40x6 integer matrix, what comes back differs from the real normalized matrix by more than 0.1 for every centering recipe. Lead with that instead, and pin both halves in tests: the identity `to_scipy_sparse().toarray() - means == toarray()` for every recipe, and the fact that dropping `means` really does change the answer for the centering ones -- so the warning cannot quietly stop being true. No API change. An earlier draft added `is_sparse` plus guarded `to_scipy()`/`to_csr()`/`to_csc()`, but `is_sparse` only restated the already-public `recipe.center`, the format helpers only restated scipy's own `.tocsr()`/`.tocsc()`, and a second materializer differing from the first only in whether it raises is more API to understand rather than less. The sparse decomposition itself landed in #47 and #51: `to_scipy_sparse()` is the `sparse_delta()` that work proposed, and `means` is its `baseline` negated. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
fishidaho
force-pushed
the
docs/to-scipy-sparse-contract
branch
from
September 18, 2026 05:03
3de005c to
7c21a4e
Compare
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
to_scipy_sparse()(#47) returns the uncenteredDeltaterm and leavesmeansfor the caller to subtract. For the two recipes that do not center that is the normalized matrix; for the three that do (parafac2,scanpy,pearson) it is not.The docstring mentioned
meansonly in passing, after describing the return value in terms that read as complete. The documented use is handing arrays to parafac2's CuPy/MLX backends, which is exactly where a silently uncentered matrix is hard to trace back.Fix
Lead the docstring with what it omits, name the recipes it bites, and give the identity for recovering the whole matrix.
Pin both halves in tests:
to_scipy_sparse().toarray() - means == toarray()for every recipe, and the fact that droppingmeansreally does change the answer for the centering ones — so the warning cannot quietly stop being true.No API change.
Measurements
On a 40x6 integer matrix,
to_scipy_sparse()differs from the true normalized matrix by more than 0.1 for every centering recipe, and is exact for the other two.Merge order
Any time, independent of the others.
🤖 Generated with Claude Code
https://claude.ai/code/session_01Rvu3cf8ZL7F5EPX22eo6Je