MCMC Lab
An interactive web application that visualizes how various Markov Chain Monte Carlo (MCMC) algorithms explore probability distributions.
Designed for students, data scientists, and researchers to gain intuition about sampling efficiency, tuning parameters, and the behavior of different samplers in 2D space.
MCMC Lab is an interactive web application that visualizes how various Markov Chain Monte Carlo (MCMC) algorithms explore probability distributions.
Here are the features for this project:
- Algorithms Implemented:
- Random Walk Metropolis (RWM)
- Independent Metropolis-Hastings
- Slice Sampling
- Elliptical Slice Sampling
- Hit-and-Run Sampler
- Hamiltonian Monte Carlo (HMC)
- Target Distributions: Gaussian, Bimodal Mixture, Donut, and Banana (Rosenbrock).
- Real-time Visualization: Watch chains evolve step-by-step on a heatmap of the target density.
- Interactive Tuning: Adjust step sizes, leapfrog steps, and speed/epsilon on the fly.