tfp.experimental.mcmc.augment_with_state_history

Decorates a transition or proposal fn to track state history.

tfp.experimental.mcmc.augment_with_state_history(
    fn
)

For example usage, see tfp.experimental.mcmc.augment_prior_with_state_history.

Args
`fn`	Python `callable` to wrap, having signature `new_state_dist = fn(step, state_with_history, **kwargs)` where `state_with_history` is a `StateWithHistory` namedtuple.

Returns
`augmented_fn`	Python `callable` wrapping `fn`, having signature `new_state_with_history_dist = augmented_fn(step, state_with_history, **kwargs)`. The return value is a `tfd.JointDistributionNamed` instance over`tfp.experimental.mcmc.StateWithHistory` namedtuples, in which the `state_history` component is rotated to discard the (previously-oldest) state at the initial position and append the new state at the final position.

Returns

augmented_fn Python callable wrapping fn, having signature new_state_with_history_dist = augmented_fn(step, state_with_history, **kwargs). The return value is a tfd.JointDistributionNamed instance overtfp.experimental.mcmc.StateWithHistory namedtuples, in which the state_history component is rotated to discard the (previously-oldest) state at the initial position and append the new state at the final position.

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2023-11-21 UTC.

English
中文 – 简体

tfp.experimental.mcmc.augment_with_state_history

Args

Returns