Aurora-GLM Documentation

A modular, extensible, and high-performance Python framework for Generalized Linear Models (GLM), Generalized Additive Models (GAM), and Generalized Additive Mixed Models (GAMM).

User Guide

Step-by-step tutorials covering GLM, GAM, GAMM, formula syntax, multi-backend support, and inference.

User Guide
API Reference

Complete reference for all public modules, classes, and functions.

API Reference
Examples

17 case studies including insurance pricing, species distribution, clinical trials, and more.

Examples & Case Studies

Features

  • Multi-backend: NumPy, PyTorch, and JAX with automatic detection

  • GPU acceleration: Up to 141x speedup via PyTorch CUDA

  • 10+ distribution families: Gaussian, Poisson, Binomial, Gamma, Beta, and more

  • 9 link functions: Logit, Log, Probit, CLogLog, and more

  • R-compatible formulas: y ~ s(x1) + x2 + (1 | group)

  • Validated against R: Results match mgcv, lme4 within 1e-6 tolerance

  • Sparse matrix support: Efficient for large n or many smooth terms

Quick Start

import aurora
from aurora import fit_glm, GaussianFamily, LogitLink

# Fit a GLM
result = fit_glm(X, y, family=GaussianFamily())

# Inspect results
result.summary()
result.plot_diagnostics()

# Use PyTorch backend for GPU acceleration
import torch
result = fit_glm(torch.tensor(X), torch.tensor(y), family=GaussianFamily())

Indices