Go Back: Hugh McDougall Astro Blog Return to Blog Home
Comfortably NumPyro
Navigation
- Getting Started - An absolute beginners guide to NumPyro
- The Constrained Domain - Explanation of constranied vs unconstrained domains in NumPyro
- Testing Different MCMC Samplers - A quick overview of the different samplers in numpyro and their strengths / weaknesses
- Nested Sampling - An exploration of nested sampling in NumPyro
- Stochastic Variational Inference - An intro and guide to SVI in NumPyro
- SVI Part 1 - Introduction & Explanation
- SVI Part 2 - Examples & Comparison with MCMC
It is a generally accepted fact amongst most reasonable people that Bayesian analysis is the correct approach to most any non trivial statistical problem. In the modern world, we enjoy with a suite of tools that alleviate the tedium of constructing and running these models: Bayesian analysis lets us do things right, probabalistic programming languages (PPLs) let us do it easily, and the magical power of Just-In-Time (JIT) compiled languages allows us to to it fast. The JAX-based PPL NumPyro brings all three together: a python interface that gives great speed and versatility, and makes ubiquitous tasks like parameter constraint or model comparison cheap in both human-time and machine-time.
For the experienced user, the interaction cost hurdle between having an idea and getting a nice ChainConsumer corner plot has never been shorter. The only issue is that some corners of NumPyro can be opaque to the unfamilar user. In this blog, I provide a handful of short and to-the-point tutorials that walk the new user through their first steps into the world of NumPyro, and guide the “almost new” user through the less-obvious features that might otherwise cost hours of scanning documentation.
If you’re already confident, you might also consider Dan Foreman Mackey’s Astronomer’s Guide to NumPyro, dive right into the extensive examples provided by NumPyro’s documentation itself.
This page by Hugh McDougall, 2024
For more detailed information, feel free to check my GitHub repos or contact me directly.