An animated wrapping of Dan Foreman-Mackey's corner.py.
prism
is handy for watching ensembles burn in. Pretend below that
samples
are samples in 5 dimensions collected by 5000 walkers in 100 steps.
import numpy as np
import prism
nsteps, nwalkers, ndim = 100, 5000, 5
samples = np.random.randn(nsteps * nwalkers * ndim).reshape([nsteps, nwalkers, ndim])
# Make-believe burn in
samples[:nsteps/10] *= np.arange(1, nsteps/10+1)[::-1, np.newaxis, np.newaxis]
anim = prism.corner(samples)
anim.save("prism.mp4", fps=30)
and you get something like