Sequince examples

Start with the basic examples to learn how to define a prior and likelihood, initialize particles, run Sequential Monte Carlo (SMC), and interpret weighted posterior samples and evidence estimates. Work through the Gaussian target first, then try the curved Rosenbrock posterior.

Getting started

• Gaussian posterior
• Rosenbrock posterior

Once you are comfortable with that workflow, explore ways to adapt mutation, change the coordinates used by the sampler, and reuse particle populations. Each example has its own tutorial page with explanations and script source. Example plots appear directly on that page.

Sampling methods

• Adaptive mutation steps
• Flow preconditioning
• Persistent sampling
• Waste-free sequential Monte Carlo

For sequential observations, the particle-filtering example compares bootstrap and observation-conditioned proposals with a Kalman-filter reference.

Sequential data

• Particle filtering with Sequince

Use the array backend examples to run differential evolution with NumPy, CuPy, or PyTorch, and to compile an SMC iteration with JAX’s functional API.

Array backends

• Array backends
  • Using array backends
  • The Himmelblau examples

The basic, adaptive mutation, flow preconditioning, persistent sampling, and waste-free SMC, and particle-filtering tutorials include plots alongside their explanations. Their run metadata identifies the source revision, parameters, seed, and pinned dependencies.

When using these examples in research, follow the software and algorithm citation guidelines for Sequince and any additional packages and methods used.

For background and the full interface reference, see Sequince’s user guide and API reference. The examples use ParticleState to represent the weighted particle population.