Come analyze HEASARC, IRSA, and MAST data in the cloud! The Fornax Initiative is now welcoming all interested beta users.
Hamiltonian Monte Carlo.
No proposal command – HMC derives its trajectories from gradients of the log-posterior. We use a deliberately small budget here to keep the example short; for a production run aim for at least 1000 warmup + 1000 samples per chain.
hmc warmup 100 hmc samples 100 hmc chains 4 hmc target_accept 0.8 parallel hmc 4 hmc run hmc.fits
Observed: 590 s elapsed time. At this small budget HMC has not yet built up a good ESS; the same model at the documented default budget (1000+1000) finishes in roughly 1.5–2 hours elapsed time and converges cleanly.