ConceptPythonAdvanced
Hamiltonian Monte Carlo: why gradients beat a random walk in 100 dimensions
- tau = integrated_time(chain[:, None, :1], quiet=True)[0]
- tau = integrated_time(run["chain"][:, None, 49:50], quiet=True)[0]
- tau_rw = THIN * integrated_time(rw[:, None, :], quiet=True)
- tau_hmc = integrated_time(hm["chain"][:, None, :], quiet=True) * hm["work"][-1] / N_HMC
- ess = n / (scale * integrated_time(y[:, None, :], quiet=True))