autointent.nodes.NodeOptimizer#

class autointent.nodes.NodeOptimizer(node_type, search_space, target_metric, metrics=None)#

Class for optimizing nodes in a computational pipeline.

This class is responsible for optimizing different modules within a node using various search strategies and logging the results.

Parameters:

• node_type (autointent.custom_types.NodeType)

• search_space (list[dict[str, Any]])

• target_metric (str)

• metrics (list[str] | None)

node_type#

node_info#

target_metric#

metrics = None#

modules_search_spaces#

fit(context, sampler='brute')#

Performs the optimization process for the node.

Parameters:

• context (autointent.context.Context) – The optimization context containing relevant data.

• sampler (autointent.custom_types.SamplerType) – The sampling strategy used for optimization.

Raises:

AssertionError – If an invalid sampler type is provided.

Return type:

None

objective(trial, module_name, search_space, context)#

Defines the objective function for optimization.

Parameters:

• trial (optuna.trial.Trial) – The Optuna trial instance.

• module_name (str) – The name of the module being optimized.

• search_space (dict[str, ParamSpaceInt | ParamSpaceFloat | list[Any]]) – The parameter search space.

• context (autointent.context.Context) – The execution context.

Returns:

The value of the target metric for the given trial.

Return type:

float

suggest(trial, search_space)#

Suggests parameter values based on the search space.

Parameters:

• trial (optuna.trial.Trial) – The Optuna trial instance.

• search_space (dict[str, Any | list[Any]]) – A dictionary defining the parameter search space.

Returns:

A dictionary containing the suggested parameter values.

Raises:

TypeError – If an unsupported parameter search space type is encountered.

Return type:

dict[str, Any]

get_module_dump_dir(dump_dir, module_name, j_combination)#

Creates and returns the path to the module dump directory.

Parameters:

• dump_dir (pathlib.Path) – The base directory for storing module dumps.

• module_name (str) – The name of the module being optimized.

• j_combination (int) – The combination index for the parameters.

Returns:

The path to the module dump directory.

Return type:

str

validate_nodes_with_dataset(dataset, mode)#

Validates nodes against the dataset.

Parameters:

• dataset (autointent.Dataset) – The dataset used for validation.

• mode (autointent.custom_types.SearchSpaceValidationMode) – The validation mode (“raise” or “warning”).

Raises:

ValueError – If validation fails and mode is set to “raise”.

Return type:

None

validate_search_space(search_space)#

Check if search space is configured correctly.

Parameters:

search_space (list[dict[str, Any]])

Return type:

None