Skip to content

RELEASE_NOTES.md

Latest commit

 

History

History
86 lines (67 loc) · 3.35 KB

RELEASE_NOTES.md

File metadata and controls

86 lines (67 loc) · 3.35 KB
Raw

Release Notes

v0.1.0 - GitHub Library Pre-Release

Status: GitHub-only library pre-release.

jet_simplex_search is a Python package for bounded directed simplex search in sparse graph data. It implements a focused release slice of Abdullah N. Malik's simplicial graph-search program: directed graph data, first-class degenerate simplices, quotient-tower search, and fiber-addressed lifting.

This release is not published to PyPI. Install from a source checkout:

git clone https://github.com/TYLERSFOSTER/jet_simplex_search.git
cd jet_simplex_search
uv sync

The package depends on state_collapser through the GitHub release tag v0.7.2:

"state-collapser @ git+https://github.com/TYLERSFOSTER/state_collapser.git@v0.7.2"

Implemented

  • Sparse graph input records.
  • H-to-G skeletonization for input loops and parallel edges.
  • Formal identity search edges for degenerate skeleton addresses.
  • Directed flag simplex enumeration through a user-provided k.
  • First-class degenerate simplex records.
  • Cached sparse frontier intersections.
  • Static quotient-tower integration through state_collapser.
  • Fiber-addressed lifting over existing downstairs simplex records.
  • Regression coverage for the current small-object behavior: if a downstairs simplex is absent, upstairs lifting over that simplex address does not run.
  • Witness consistency checks for lifted simplex records.
  • Compressed H-lift counts for tier-0 skeleton simplices.
  • JSON and JSONL artifact output.
  • README quick-start regression test.
  • Low-dimensional smoke examples through dimension 4.

Deferred

  • Kan replacement and horn-filling variants.
  • Full cofibrant or small-object replacement API.
  • Expanded H witness assignment artifacts.
  • Bitset, CSR, GPU, tensor, or multiprocessing acceleration.
  • Dynamic tower rebuilding during simplex search.
  • PyPI publication.
  • No benchmark-validated speed-up claims are made.

This package also does not implement neural message passing, CinchNET, or PTVNN. The separate HGraphML package is the beginning of the quotient-tower-backed graph message-passing branch; jet_simplex_search is the simplex-search and quotient-lift branch.

Verification

Recommended local checks:

uv run ruff check .
uv run ruff format --check .
uv run pytest
uv run python scripts/release_hygiene.py --repo-root .
uv build --out-dir <temporary-build-directory>

Design Record

Attribution

The mathematical and algorithmic origin of this package is Abdullah N. Malik's work on simplicial methods in graph machine learning, especially the quotient-tower simplex search line described in his thesis and related project work. This package is a release-oriented implementation track around that lineage and its state_collapser integration.