pypomp documentation

Version: 0.4.6.0 Date: June 05, 2026

pypomp is a Python package for modeling and inference using partially observed Markov process (POMP) models, also called state-space models (SSM) or hidden Markov models (HMM). Key features include:

• Estimation, filtering, and inference for nonlinear, non-Gaussian POMP models via the particle filter

• GPU support and just-in-time compilation via JAX, enabling significant speedups

• New algorithms for model-fitting with gradient descent using improved gradient estimates

Installation

You can install pypomp from PyPI:

pip install pypomp

pypomp depends on JAX. To take full advantage of GPU acceleration, we highly recommend installing the GPU-enabled version of JAX. Please refer to the JAX installation guide for detailed instructions specific to your system.

Getting started

• The tutorials provide introductory explanations of POMP models and methods in pypomp.

• The quantitative tests provide sample code used for benchmarking package performance.

• The pypomp organization GitHub site hosts source code and related projects.

The main classes in pypomp are:

• Pomp - Core POMP model class

• PanelPomp - Panel POMP models for multiple units

Contents

Getting Started

• Tutorials
• Model Structure Components
  • State Initialization (rinit)
  • State Transition (rproc)
  • Measurement Density (dmeas)
  • Measurement Simulator (rmeas)
• Best Practices and Performance Guidelines
  • Essential Guidelines

Reference

• API Reference
  • Core Classes
  • Model Components
  • Utilities
  • Example Models

Resources

• Quantitative Tests
• GitHub

Indices

• Index

• Module Index

• Search Page