PEA-PGNN v0.1.2

Maintenance release that removes an unused comparison utility and its documentation.

The concrete prior anchor continues to use Model B3, ACI 209, and GL2000. The four candidate temporal laws remain unchanged.

Install or upgrade with:

python -m pip install --upgrade pea-pgnn

PEA-PGNN v0.1.1

PEA-PGNN 0.1.1 is the first PyPI-ready release of the reusable prior-anchored, structure-preserving neural prediction package.

Highlights:

• documented empirical-prior provenance, units, simplifications, and applicability limits;
• added a model card, security policy, contribution guidance, Code of Conduct, and issue/PR templates;
• expanded tests for concrete utilities, metrics, invalid inputs, EC2 behavior, and NumPy/PyTorch parity;
• added coverage enforcement and multi-version CI for Python 3.9, 3.11, and 3.13;
• added OIDC-based Trusted Publishing workflows for TestPyPI and PyPI;
• aligned package, citation, author, and release metadata at version 0.1.1.

The package contains reusable method code and concrete empirical-prior utilities. It does not contain the manuscript database, trained weights, or submission files.

Install after the PyPI publication completes:

python -m pip install pea-pgnn

See CHANGELOG.md for the full list of changes.

PEA-PGNN v0.1.0

PEA-PGNN v0.1.0

Initial open-source release of the prior-anchored, structure-preserving neural prediction library.

Included

• Prior-anchored PyTorch temporal model
• B3-, GL2000-, ACI209-, and EC2-inspired concrete shrinkage utilities
• Four normalized temporal candidate laws and convex mixing
• NumPy-facing training, prediction, save, and load interface
• Numerical audits for non-negativity, monotonicity, and boundedness
• Documentation, synthetic example, citation metadata, and automated tests

Installation

Install the attached wheel:

python -m pip install pea_pgnn-0.1.0-py3-none-any.whl

Or install directly from GitHub:

python -m pip install "git+https://github.com/hunter137/pea-pgnn.git@v0.1.0"

The manuscript database and submission materials are intentionally not included.