HMCKernel

class liesel.goose.HMCKernel(position_keys, initial_step_size=None, initial_inverse_mass_matrix=None, num_integration_steps=10, da_target_accept=0.8, da_gamma=0.05, da_kappa=0.75, da_t0=10, mm_diag=True, identifier='')

Bases: ModelMixin, TransitionMixin[HMCKernelState, HMCTransitionInfo], TuningMixin[HMCKernelState, DefaultTuningInfo], ReprMixin

A HMC kernel with dual averaging and an inverse mass matrix tuner, implementing the Kernel protocol.

Parameters:

• position_keys (Sequence[str]) – Sequence of position keys (variable names) handled by this kernel.

• initial_step_size (float | None, default: None) – Value at which to start step size tuning.

• initial_inverse_mass_matrix (Any | None, default: None) – Starting value for the inverse mass matrix (the precision matrix of the momentum). If None, an identity matrix will be used here.

• num_integration_steps (int, default: 10) – Number of integration steps used in the leapfrog algorithm. Corresponds to L in the Stan reference manual [1].

• da_target_accept (float, default: 0.8) – Target acceptance probability for dual averaging algorithm.

• da_gamma (float, default: 0.05) – The adaptation regularization scale.

• da_kappa (float, default: 0.75) – The adaptation relaxation exponent.

• da_t0 (int, default: 10) – The adaptation iteration offset.

• mm_diag (bool, default: True) – Whether to use a diagonal mass matrix for drawing the momentum vector. If True, the inverse mass matrix will be tuned during adaptation using tune_inv_mm_diag(). If set to False, the mass matrix will be tuned using tune_inv_mm_full() instead.

• identifier (str, default: '') – An string acting as a unique identifier for this kernel.

Notes

For more information on step size tuning via dual averaging, see da_step() and DAKernelState.

[1]

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

Methods

end_epoch(prng_key, kernel_state, ...)	Sets the step size as found by the dual averaging algorithm.
end_warmup(prng_key, kernel_state, ...)	Currently does nothing.
init_state(prng_key, model_state)	Initializes the kernel state with an identity inverse mass matrix and a reasonable step size (unless explicit arguments were provided by the user).
start_epoch(prng_key, kernel_state, ...)	Resets the state of the dual averaging algorithm.

Attributes

error_book	Dict of error codes and their meaning.
identifier	Kernel identifier, set by EngineBuilder
needs_history	Whether this kernel needs its history for tuning.
position_keys	Tuple of position keys handled by this kernel.