DAKernelState

class liesel.goose.da.DAKernelState(*args, **kwargs)

Bases: Protocol

A protocol for a kernel state with dual averaging support. For an introduction to dual averaging, see the blog post by Colin Carroll [1] and the Stan Reference Manual [2].

[1]

Colin Carroll, Step Size Adaptation in Hamiltonian Monte Carlo (2019).

[2]

Stan Development Team, Stan Reference Manual (2021), Chapter 15.2.

Attributes

step_size	The step size of the kernel.
error_sum	The error sum of the acceptance probability.
log_avg_step_size	The logarithm of the average step size.
mu	The bias of the step size proposals.