Skip to content

# Aurora-GLM v0.7.0 - First Stable Release

Latest
Latest

Choose a tag to compare

View all tags
@Matcraft94 Matcraft94 released this 07 Dec 00:57
v0.7.0
6672719
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Aurora-GLM v0.7.0 - First Stable Release

Aurora-GLM v0.7.0 is here! The first stable release of Aurora-GLM,
featuring complete MIT License implementation, Phase 5 completion, and
production-ready statistical modeling capabilities.

Complete MIT License Implementation

  • Full LICENSE file with copyright (2025 Lucy Eduardo Arias)
  • SPDX headers on all 137 Python source files
  • CITATION.cff for academic attribution with ORCID
  • Professional authors documentation
  • Free software: unlimited use, modification, distribution with attribution

17 Comprehensive Case Studies

Production-ready examples across domains:

  • Finance & Insurance: Insurance pricing, claims, churn prediction
  • Environmental: Air quality, species distribution, bike sharing
  • Healthcare: Sleep studies, clinical trials, cancer survival
  • Education: Multilevel student achievement
  • Business: Wind power, restaurant health inspection

All follow 8-part structure: Problem → Data → Model Selection → Specification
→ Fitting → Results → Diagnostics → Conclusions

Phase 5 Features Complete

Formula Validation

  • R-style syntax validation
  • Clear error messages
  • Prevents misinterpretation

Sparse Optimization

  • 6-8× memory reduction
  • 10-100× speedup
  • Efficient EDF computation

GAMM Enhancements

  • Log-likelihood computation
  • AIC/BIC metrics
  • 8 covariance structures

Streaming Support

  • Memory-mapped arrays
  • Large file handling
  • Efficient RAM usage

Advanced Smoothing

  • P-splines (GCV/AIC/REML)
  • LOESS local regression
  • Tensor products

Extended Models

  • Zero-Inflated (ZIP, ZINB)
  • Hurdle (HP, HNB)
  • Bayesian GLM (NumPyro/PyMC)

Distributed Computing

  • SGD optimization
  • Data-parallel IRLS
  • Scalable estimation

Complete Feature Set

Distributions: Gaussian, Binomial, Poisson, Gamma, Beta, NegBin, Tweedie,
Student-t, InverseGaussian, Zero-Inflated, Hurdle variants

Links: Identity, Log, Logit, Probit, Log-log, Inverse, Sqrt, Power

Smooth Terms: B-splines, Natural cubic, Thin plate, Tensor products, LOESS

Random Effects: Intercepts, slopes, nested, crossed effects

Covariance: Identity, Unstructured, Diagonal, AR1, Compound Symmetry,
Exponential, Matérn, Toeplitz

Backends: NumPy, PyTorch (CPU/GPU), JAX (CPU/GPU)

Quality Assurance

  • 479+ unit tests with multi-backend validation
  • Accuracy: Validated against R glm() and statsmodels
  • Performance: GPU benchmarks up to 141× speedup
  • Backward compatible: No breaking changes from v0.6.1

Attribution

Lucy Eduardo Arias (non-binary, they/them)

License

MIT License - Completely free, with attribution

  • Commercial use, modification, distribution
  • Private use
  • Condition: Include license and copyright notice
  • No warranty, no liability

Getting Started

Installation (coming to PyPI)

pip install aurora-glm  # Soon
pip install -e .        # From source

Citation

@software{aurora_glm2025,
   title = {Aurora-GLM: Generalized Linear and Additive Models},
   author = {Arias, Lucy Eduardo},
   year = {2025},
   version = {0.7.0},
   url = {https://github.com/Matcraft94/Aurora-GLM},
   license = {MIT}
}

Or use GitHub's "Cite this repository" button with CITATION.cff

Documentation

🙏 Acknowledgments

Inspired by R's mgcv, lme4, Python's statsmodels, and the JAX ecosystem.

Contributors

Matcraft94
Assets 2
Loading