Repository navigation
feat(adr-001 #6): phase-2B — single-entry Neumann + SubLinear orchestrator - #27
Merged
Merged
Conversation
…rator
Lands the terminal SubLinear form of the contrastive solve. The
phase-2A orchestrator was Adaptive {Linear, Linear} because its
inner solve_on_change materialised the full solution vector even
when only |closure| ≪ n rows mattered. Phase-2B replaces that with
per-entry sublinear-Neumann queries scoped to the closure, and the
orchestrator now declares end-to-end SubLinear.
Lands:
src/entry.rs (new, 381 LOC)
solve_single_entry_neumann(matrix, b, target, max_terms, tolerance)
Computes x[i] = e_iᵀ A⁻¹ b for strict-DD A by truncated Neumann
series restricted to closure(i, max_terms). Never materialises x.
O(max_terms · |closure| · branching) — independent of n for
sparse DD matrices with bounded depth. Op marker
SolveSingleEntryNeumannOp declares SubLinear.
solve_single_entries_neumann(matrix, b, targets, max_terms, tolerance)
Batched variant returning sorted (idx, val) pairs — the natural
feed for find_anomalous_rows_in_subset.
src/contrastive.rs (+143 LOC)
contrastive_solve_on_change_sublinear(matrix, prev, b_new, delta,
closure_depth, max_terms, tolerance, k)
End-to-end SubLinear: closure (SubLinear) → per-entry Neumann
(SubLinear) → top-k-in-subset (SubLinear). Empty delta / zero-k
short-circuit before any solve work — the "no event, no work"
path is free, which is the gating discipline the ADR turns on.
Op marker ContrastiveSolveOnChangeSublinearOp declares SubLinear.
src/lib.rs
Re-exports solve_single_entry{,ies}_neumann, SolveSingleEntryNeumannOp,
contrastive_solve_on_change_sublinear, ContrastiveSolveOnChangeSublinearOp.
docs/adr/ADR-001-complexity-as-architecture.md
Item #6 status updated to "phase-2B landed — end-to-end SubLinear".
Status preamble notes that the change-driven activation primitive
is now fully realised.
.github/complexity-baseline.txt
Adds SolveSingleEntryNeumannOp (SubLinear) and
ContrastiveSolveOnChangeSublinearOp (SubLinear) so the regression
guard knows the new declarations.
Tests: 188 lib tests pass (was 176 before phase-2A; +12 across entry
+ contrastive). Critical correctness test: entry::tests::
matches_full_solve_on_chain verifies the single-entry estimate
agrees with a full NeumannSolver solve at every row of an 8×8 DD
chain within 1e-6.
Co-Authored-By: claude-flow <ruv@ruv.net>
This was referenced May 19, 2026
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.
Summary
Lands the terminal SubLinear form of the contrastive solve. Phase-2A was
Adaptive { Linear, Linear }because its innersolve_on_changematerialised the full solution vector even when only|closure| ≪ nrows mattered. Phase-2B replaces that with per-entry sublinear-Neumann queries scoped to the closure, and the orchestrator now declares end-to-endSubLinear.What's new
src/entry.rs(new) —solve_single_entry_neumann(matrix, b, target, max_terms, tolerance): computesx[i] = e_iᵀ A⁻¹ bby truncated Neumann iteration restricted toclosure(i, max_terms). Never materialisesx. CostO(max_terms · |closure| · branching)— independent ofn. Op markerSolveSingleEntryNeumannOpdeclaresSubLinear. Batched variantsolve_single_entries_neumannfor fan-out at orchestrator time.src/contrastive.rs(+143 LOC) —contrastive_solve_on_change_sublinear(matrix, prev, b_new, delta, closure_depth, max_terms, tolerance, k): end-to-end SubLinear orchestrator. Empty-delta / zero-k short-circuit before any solve work — the "no event, no work" path is free.complexity-baseline.txtregenerated.Correctness gate
entry::tests::matches_full_solve_on_chain— single-entry estimate agrees with a fullNeumannSolversolve at every row of an 8×8 strict-DD chain within1e-6.contrastive::tests::sublinear_orchestrator_finds_changed_rows_on_chain— perturbingb[2]and asking for top-3 anomalies returns a top-k that includes row 2 (the change epicentre) with anomaly scores ordered descending.Test plan
cargo test --lib→ 188/188 pass (was 176 pre-phase-2A; +12 entry + orchestrator tests)cargo check --libclean🤖 Generated with claude-flow