odatse.algorithm.bayes module#

class odatse.algorithm.bayes.Algorithm(info: Info, runner: Runner = None, domain=None, run_mode: str = 'initial')[source]#

Bases: AlgorithmBase

A class to represent the Bayesian optimization algorithm.

mesh_list#

The mesh grid list.

Type:

np.ndarray

label_list#

The list of labels.

Type:

list[str]

random_max_num_probes#

The maximum number of random probes.

Type:

int

bayes_max_num_probes#

The maximum number of Bayesian probes.

Type:

int

score#

The scoring method.

Type:

str

interval#

The interval for Bayesian optimization.

Type:

int

num_rand_basis#

The number of random basis.

Type:

int

xopt#

The optimal solution.

Type:

np.ndarray

best_fx#

The list of best function values.

Type:

list[float]

best_action#

The list of best actions.

Type:

list[int]

fx_list#

The list of function values.

Type:

list[float]

param_list#

The list of parameters.

Type:

list[np.ndarray]

Constructs all the necessary attributes for the Algorithm object.

Parameters:

  • info (odatse.Info) – The information object.

  • runner (odatse.Runner, optional) – The runner object (default is None).

  • domain (optional) – The domain object (default is None).

  • run_mode (str, optional) – The run mode (default is “initial”).

__init__(info: Info, runner: Runner = None, domain=None, run_mode: str = 'initial') → None[source]#

Constructs all the necessary attributes for the Algorithm object.

Parameters:

  • info (odatse.Info) – The information object.

  • runner (odatse.Runner, optional) – The runner object (default is None).

  • domain (optional) – The domain object (default is None).

  • run_mode (str, optional) – The run mode (default is “initial”).

_apply_state(data: dict, mode: str = 'resume', restore_rng: bool = True) → None[source]#

Restore algorithm state from a checkpoint snapshot.

Delegates MPI validation, timer restore, parameter check, and the instance RNG restore to the base class; additionally restores the global numpy RNG used by physbo, applies the Bayes-specific fields, then loads the physbo policy from the saved files.

_load_state() is not overridden; the base class implementation calls self._apply_state() (this method), so policy files are loaded automatically when resuming from a checkpoint.

Parameters:

  • data (dict) – Snapshot previously produced by __getstate__.

  • mode (str) – "resume" or "continue"; forwarded to the base class. Bayes optimisation does not distinguish between the two modes.

  • restore_rng (bool) – When True (default) both RNG states are restored from data.

_initialize()[source]#

Initializes the algorithm parameters and timers.

_post() → dict[source]#

Finalizes the algorithm execution and writes the results to a file.

_prepare() → None[source]#

Algorithm-specific preparation, called after dispatch.

Override in subclasses to perform setup that must happen after the checkpoint state is established (e.g. initialising timer entries).

_run() → None[source]#

Runs the Bayesian optimization process.

_save_state(filename)[source]#

Save the current state of the algorithm to a file.

Extends the base implementation with physbo policy files. The pickle snapshot (including the global numpy RNG captured in __getstate__) is written by super()._save_state().