tutorials

Pypomp Tutorials & Courses

Here, we collect instructional materials to help you build models and perform inference using Pypomp, a Python/JAX library for data analysis using partially observed Markov process (POMP) models. Pypomp emphasizes the use of sequential Monte Carlo “particle filter” techniques, GPU acceleration, and automatic differentiation.

🚀 Quick Links

• 📖 Read the Documentation — Core library API reference and user guide.
• 💻 Code Repository — Source code and issue tracker for the pypomp package.
• 🎓 Tutorials Repository — Quarto notebooks and materials for learning Pypomp.

---

📚 Tutorial & Course Catalog

1. Introduction to Pypomp

A step-by-step introduction demonstrating how to construct a POMP model (specifically, an AR(1) process observed with noise) from scratch. It guides you through simulation, estimating log-likelihoods with the particle filter, and running iterated filtering (IF2) for parameter estimation, checking the results against the exact answers from the Kalman filter.

2. Simulation-based Inference for Epidemiological Dynamics (SBIED)

A short course designed for readers interested in POMP modeling, using epidemiological dynamics as an example. The course transfers concepts from the R-pomp short course to Pypomp.

Course Lessons:

• Lesson 0: Preparing your laptop environment
• Lesson 1: Introduction: What is SBIED? POMPs and Pypomp
• Lesson 2: Simulation of stochastic dynamic models
• Lesson 3: Likelihood for POMPs: theory and practice
• Lesson 4: Iterated filtering: theory and practice
• Lesson 5: Case study: Measles (recurrent epidemics, covariates)
• Lesson 6: Case study: Ebola (model diagnostics and forecasting)

3. Inference for Cholera Dynamics in Dhaka using DMOP

A tutorial demonstrating how to construct a POMP model for cholera dynamics in Dhaka from scratch and how to perform parameter estimation using the Differentiated Measurement Off-Parameter (DMOP) filter with the Adam optimizer.

4. Fitting a Large Panel Model for UK Measles

A tutorial demonstrating how to construct and fit a large spatial panel model (with weekly case reports from 1,422 cities/towns in England and Wales) using Marginalized Panel Iterated Filtering (MPIF).

5. Fitting a Four Species Ecological Model to Panel Data

A tutorial demonstrating how to construct and fit a panel model for a system of four interacting species, using Marginalized Panel Iterated Filtering (MPIF).

---

💬 Community and Support

• Docs: Refer to the core Read the Docs page.
• Bugs & Features: Please open an issue or submit a pull request on our main GitHub repository.
• Contact: Pypomp is under active development. Please get in touch if you are reading this and you have feedback! You can email the core team at pypomp-org@umich.edu or reach out via GitHub.

This site is open source. Improve this page.