Politics & Elections
Voting patterns, electoral forecasting, and spatial models for political behavior.
Application areas
INLA is used across fields where uncertainty quantification and spatial-temporal modeling are essential.
Business & Finance
Demand forecasting, credit risk, and regional economic performance.
Public Health
Disease mapping, mortality trends, and outbreak detection.
Probabilistic ML
Gaussian processes, state-space models, and structured random effects with uncertainty.
Urban & Transport
Mobility patterns, traffic flow, and demand modeling.
Sports Analytics
Player performance, win probability, and expected goals (xG).
Environment & Climate
Species distributions, air quality, and climate impact assessment.
Spatial & Geostatistics
Point patterns, areal data, and continuous spatial fields.
Why pyINLA?
Native Python access to INLA with intuitive syntax, pandas integration, and built-in diagnostics.
Performance
INLA speed
Full posterior inference in seconds through the INLA computational engine.
Syntax
Formula interface
Specify models with fixed effects, random effects, and spatial components using familiar syntax.
Diagnostics
Built-in checks
Posterior summaries, PIT histograms, and residual diagnostics included.
Integration
pandas compatible
Works directly with DataFrames and integrates with matplotlib, seaborn, and other Python tools.
Extensibility
Modular design
Add custom likelihoods and latent effects through documented extension points.
Documentation
Worked examples
Step-by-step tutorials, API reference, and reproducible notebooks.
Where pyINLA fits
Decision infrastructure for domains where uncertainty matters.
- Fast Bayesian inference for Latent Gaussian Models
- Calibrated uncertainty for decision-making in policy, health, energy, and risk
- Spatial, temporal, and spatio-temporal modeling at scale
Suitable for research and early applied deployments where calibrated uncertainty and fast iteration are required.
News & Updates
Stay up to date with the latest developments in pyINLA.
Status update
Testing phase underway
pyINLA is currently in an internal testing phase while we validate the Python interface. Expect rapid changes as we harden the API, broaden likelihood coverage, and publish full documentation.
What to expect
Over the next releases we will continue porting worked examples, exercising every feature against the reference workflows, and publishing migration notes. Follow this space for announcements as additional likelihoods, latent models, and notebooks graduate from testing to general availability.
Get involved
Want to receive updates as new features are released, or interested in contributing to a specific likelihood or latent model? Contact us at contact@pyinla.org and let us know what you'd like to work on or follow.
Streamlined workflow
pyINLA provides sensible defaults while allowing explicit configuration when needed. The same workflow applies to spatial, temporal, and survival models.
- Type-annotated API with informative error messages
- Reproducible results with versioned outputs
- Export to CSV, JSON, and other standard formats
What is INLA?
The Integrated Nested Laplace Approximation (INLA) is a strategy for fast, approximate Bayesian inference in Latent Gaussian Models, enabling rich models with practical runtime.
How INLA works ↗Ready to build your model?
Install pyINLA and follow the quickstart to fit your first model.
pip install pyinlaWe'd love to hear from you
Found a bug? Have a feature request? Want to share feedback or ask a question? Get in touch and let us know.