probml/dynamax

Health v2 · maintenance only

Adoption not part of the score

1010 stars · 114 forks observed · 2026-09-25

What it is AI-extracted, prompt v1, taxonomy v1, 2026-09-07, confidence 0.95

Dynamax is a Python library for probabilistic state space modeling built on JAX. It provides inference and learning algorithms for HMMs, linear Gaussian, nonlinear, and generalized Gaussian state space models.

Use cases

• fit hidden markov models to sequence data
• run kalman filtering and smoothing on time series
• learn parameters of linear dynamical systems
• bayesian inference for state space models with jax
• do state estimation in nonlinear ssm models
• estimate ssm parameters with sgd or hmc

When to choose

• you need probabilistic state space models with jax jit speed and autodiff
• you want a well-tested open-source library covering HMMs and Kalman filtering in one place
• you want integration with the JAX ecosystem like optax and blackjax

When to avoid

• you need deep learning sequence models like RNNs or transformers
• you prefer PyTorch or TensorFlow rather than JAX
• you need production deployment pipelines rather than a research library

Facets

library · maturity active

machine-learning math data-science machine-learning data-science time-series python jvm cross-platform state-space-models hidden-markov-models kalman-filter jax bayesian-inference time-series algorithms

Member repositories

For agents

markdown · JSON · MCP: product_card(name="probml/dynamax")