The Sampling Gallery

Launch the tool

The Sampling Gallery is an interactive, browser-based collection of Monte Carlo, MCMC, and related sampling algorithms. On paper, each algorithm is only a few lines of mathematics, but its behaviour is not always obvious. These visualisations let you watch a sampler explore a landscape in real time — where it proposes, what it rejects, and how gradients, trajectories, and adaptation shape the search. Pick a target distribution — a banana, a ring, a mixture of hills — choose an algorithm, tune its parameters, and watch it run.

The tool is designed for teaching and for building intuition about how different samplers behave — where they mix well, where they struggle, and how their tuning parameters affect performance. It builds on the excellent mcmc-demo by Chi Feng.

Samplers

The gallery currently includes the following algorithms:

Independent Monte Carlo

  • Rejection Sampling
  • Importance Sampling (with optional SIR resampling and Pareto-smoothed importance sampling)
  • Quasi-Monte Carlo

Markov chain Monte Carlo

  • Random Walk Metropolis
  • Adaptive Metropolis
  • Hamiltonian Monte Carlo
  • No-U-Turn Sampler (NUTS)
  • Riemannian-Manifold HMC
  • Apogee-to-Apogee Path Sampler
  • Metropolis-adjusted Langevin (MALA)
  • Slice Sampling
  • Elliptical Slice Sampling
  • Gibbs Sampling
  • Hessian-Hamiltonian Monte Carlo (H2MC)
  • Unadjusted Langevin (ULA)
  • Stochastic Gradient Langevin Dynamics (SGLD), with optional control variates
  • Tuning-free ULA (FUSE)

Ensemble and multimodal methods

  • Ensemble MCMC (affine-invariant stretch moves, with a differential evolution mode)
  • Parallel Tempering

Sequential and annealed methods

  • Tempered Sequential Monte Carlo (SMC), with an annealed importance sampling mode
  • Nested Sampling

Non-reversible samplers

  • Zig-Zag Sampler
  • Bouncy Particle Sampler

Deterministic and variational particle methods

  • Stein Variational Gradient Descent
  • Coin SVGD
  • Wasserstein Particle Descent
  • Stochastic Particle-Optimization Sampling

Ask a question or open an issue

The source code is available on GitHub . Suggestions for new samplers or target distributions are welcome — please open an issue on the repository.