v0.1 - Python projection release

Python 0.1 provides the Frobenius projection onto matrix Ky Fan k-norm dual balls and the corresponding proximal mapping.

• General k: a safeguarded threshold solver with active-set verification.
• Specialized k=2: an independent implementation using at most two simplex projections.
• Real and complex rectangular input, plus an explicit Hermitian/eigh path. No numerical rank truncation.
• Original MATLAB source snapshots, a Python baseline, 23 tests, four example configurations, replayable outputs, and complete benchmark records.
• The complete ZIP includes NumPy 2.3.5 for offline installation with CPython 3.12 on Windows x86-64. Install Python separately. Online installation is provided for other supported platforms; those platforms have not been tested.

Extract the complete ZIP and run these commands from its directory on Windows:

py -3.12 install.py
.\.venv\Scripts\python.exe examples\demo.py
.\.venv\Scripts\python.exe -I tools\validate.py

Assets:

• knorm_projection-0.1.zip: complete source, installer, examples, and offline dependencies.
• knorm_projection-0.1.zip.sha256: SHA-256 checksum for the ZIP.
• knorm_projection-0.1-py3-none-any.whl: the Python package for an existing compatible environment.

Clean offline installation, all 23 tests, four configurations, and fixed-input replay were verified. Benchmarks cover the recorded Python synthetic cases; some small or rectangular cases did not improve. The measurements do not establish MATLAB or outer-solver speedups.

This English documentation revision refreshes the v0.1 source tag and assets. Public documentation, documentation examples, and wheel metadata are in English. The algorithm source, original MATLAB snapshots, mathematical model, tolerances, and recorded numerical results are unchanged. Earlier local archives are preserved; use the checksum accompanying the current assets.

Installation and API · Algorithm and benchmarks