v0.0.6

What's Changed

• ci: SHA-pin GitHub Actions to immutable commits (supply-chain hardening) by @smaniches in #58
• Consistency and metadata cleanup: reconcile claim counts, fix stale docstring, remove dropped service scaffolding by @smaniches in #59
• chore(release): cut v0.0.6 by @smaniches in #60
• chore: add Dependabot config and Code of Conduct by @smaniches in #61
• fix: silence torch sparse-tensor UserWarning in sparse_scipy_to_torch by @smaniches in #67
• docs(changelog): fold the #67 warning fix into the 0.0.6 release notes by @smaniches in #68
• ci: auto-tag releases on version-bump merge to main by @smaniches in #69

Full Changelog: v0.0.5...v0.0.6

v0.0.5

What's Changed

• Fix Hodge isolated-simplex normalization and Rips H0 truncation, add cubical gradcheck by @smaniches in #56
• chore(release): cut v0.0.5 by @smaniches in #57

Full Changelog: v0.0.4...v0.0.5

v0.0.4

What's Changed

• docs(stats): line 33 - per-hyp vs global BH are a 1-for-1 swap, not nested by @smaniches in #39
• chore+docs: version reconciliation (dynamic 0.0.3) + metadata polish + doc-honesty (supersedes #40) by @smaniches in #33
• docs: fix Pages 404 links, serve sitemap.xml, complete post-hoc design timeline by @smaniches in #42
• docs(stats): make §2 key-claims table reproduce + fix per-family vs investigation-wide label by @smaniches in #43
• fix(techdebt): float64-safe sparse conversion + general-case PSD test + doc count/version by @smaniches in #44
• stats: investigation-wide FDR over 59 distinct comparisons (primary), 76 reported alongside by @smaniches in #45
• stats: demote topology-divergence to exploratory (floor-limited, no control) by @smaniches in #46
• docs: reconcile remaining 59-distinct stragglers (37%->47%/38%; LIMITATIONS/REVIEWER 76->59) by @smaniches in #47
• fix: Windows-runnable headline + reviewer commands, negative-finding reframe, version sync by @smaniches in #48
• chore: excellence QA pass (de-emoji, coverage-claim honesty, stale-test + mypy fixes) by @smaniches in #49
• ci: full-deps coverage gate; close package branch coverage to 100/100 by @smaniches in #50
• docs: state genuine 100% line+branch package coverage and the CI gate by @smaniches in #51
• chore: reconcile divergence-claim to exploratory, gate repro-script verdict, add discoverability URLs + runtime-introspectable type hints by @smaniches in #52
• ci: mark Run Experiment as manual, non-gating research job by @smaniches in #53
• docs(problems): add persistent self-audit log (PROBLEMS.md) by @smaniches in #54
• chore(release): cut v0.0.4 (ship post-0.0.3 hardening; adopt concept DOI) by @smaniches in #55

Full Changelog: v0.0.3...v0.0.4

v0.0.3

What's Changed

• ci: update experiment workflow defaults to COLLAB L_1 configuration by @smaniches in #26
• fix: cast seeds input to int in experiment workflow by @smaniches in #27
• docs: investigation-wide statistical summary (76 comparisons, global FDR) by @smaniches in #29
• fix: cast epochs input to int in experiment workflow by @smaniches in #28
• ci: enforce mypy strict and coverage gate; add REVIEWER.md and CLAIMS_TO_EVIDENCE.md by @smaniches in #30
• ci: exclude COLLAB from benchmark smoke tier by @smaniches in #31
• docs: add STATUS.md and update H011-b compute-constraint status by @smaniches in #32
• docs: align public framing with the investigation's negative primary finding by @smaniches in #34
• docs: add GitHub Pages documentation site (just-the-docs) by @smaniches in #35
• chore: add codemeta.json and topological-data-analysis keyword by @smaniches in #36
• fix(stats): correct off-by-one BH / non-significant counts in STATISTICAL_SUMMARY by @smaniches in #37
• ci: PyPI Trusted Publishing release workflow (+ bump to 0.0.3) by @smaniches in #38

Full Changelog: v0.0.2...v0.0.3

v0.0.2

What's Changed

• fix(ci): unblock test/lint matrix on this branch by @smaniches in #1
• docs: align README badge, test count, and citation with reality by @smaniches in #2
• benchmarks: rigorous comparison framework for differentiable PH (Phase 1) by @smaniches in #3
• benchmarks: add Hodge-subsystem bench (HodgeMP on MUTAG) by @smaniches in #6
• notebooks: re-land PR #7 onto main (notebooks/ + scripts/ never propagated) by @smaniches in #8
• nn: differentiable cubical PH + DRIVE U-Net topology-loss training pipeline by @smaniches in #9
• benchmarks/stats: BCa + block-bootstrap intervals + property-based tests by @smaniches in #10
• training: empirical proof that topology detects divergence ≤ a val-loss watchdog by @smaniches in #11
• gemini-backlog: address open review comments from PRs #1, #4, #5, #6, #7, #8 by @smaniches in #12
• benchmarks: add --quick CI tier so workflows fit the 30-min timeout by @smaniches in #13
• docs+empirical: MUTAG negative result (HodgeMP minimal < MLP, p_BH = 5.66e-05); README and LIMITATIONS rewrite to back every claim with citation or in-repo experiment by @smaniches in #14
• hodge/proteins: hypothesis 002 resolved — two-dataset equality holds; H6 strict-positive refuted; CI false-positive fix bundled by @smaniches in #16
• hodge/ablation: symm-normalised Hodge MP matches MLP on MUTAG (p_BH=0.714, median Δ=+0.000); positive equality claim established by @smaniches in #15
• benchmarks/cli: also exempt UnavailableBackend; surface failing cells on stderr by @smaniches in #17
• add LEADERBOARD.md + close 14 strict-mypy errors (drop continue-on-error) by @smaniches in #18
• CHANGELOG: prepare v0.0.2 entry — release gate met (NCI1 strict positive), tag pending user approval by @smaniches in #20
• hodge/nci1: HYPOTHESIS 003 RESOLVED — first strict positive-difference claim. Hodge + residual BEATS MLP on NCI1 at p_BH=4.83e-03 (+8.6 pp) by @smaniches in #19
• hodge/H004: sample-size mechanism REFUTED, H003 reproduced; H005 (feature-density) infrastructure + background run by @smaniches in #21
• Add H006 artifact-backed resolver and tests by @smaniches in #22
• H007: graph-structural signal decomposition (analysis-only) by @smaniches in #23
• v0.0.2: Academic archival with research report and citation metadata by @smaniches in #24
• Remove draft architecture docs; fix stale README text by @smaniches in #25

Full Changelog: v0.0.1...v0.0.2