colda.algorithm.train_stage package

Submodules

colda.algorithm.train_stage.api module

class colda.algorithm.train_stage.api.MakeResidual

Bases: BaseAlgorithm

Calculate Residual. Residual is used as target for the following training step.

Attributes

None

Methods

make_residual

classmethod compute_residual(task_mode: Literal['classification', 'regression'], output: ndarray, target: ndarray) → ndarray

Compute residual based on task mode, output and target

Parameters

task_mode : str target : list[str]

Returns

np.ndarray

classmethod make_init(task_mode: Literal['classification', 'regression'], target: list[str]) → ndarray

Create initial residual at round 1

Parameters

task_mode : str target : list[str]

Returns

np.ndarray

classmethod make_residual(round: str, dataset_path: str, target_idx: str, skip_header: str, task_mode: Literal['classification', 'regression'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC'], sponsor_matched_identifers: dict[str, Any], last_round_result: None) → ndarray

classmethod make_residual(round: str, dataset_path: str, target_idx: str, skip_header: str, task_mode: Literal['classification', 'regression'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC'], sponsor_matched_identifers: dict[str, Any], last_round_result: Any) → ndarray

Calculate Residual.

Parameters

round : str dataset_path : str target_idx : str skip_header : str task_mode : str metric_name : str sponsor_matched_identifers : dict[str, Any] last_round_result : Union[Any, None]=None

Returns

tuple[np.ndarray[Any], dict[str, Any]]

class colda.algorithm.train_stage.api.MakeResult

Bases: BaseAlgorithm

Combine the sponsor’s trained model and assistors’ trained models to a better sponsor model

Methods

make_result

classmethod make_result(user_id: str, train_id: str, rounds, dataset_path, target_idx, skip_header, task_mode: Literal['classification', 'regression'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC'], sponsor_trained_cooperative_model_output, assistor_trained_cooperative_model_outputs, sponsor_matched_identifers, last_round_result) → None

classmethod result_func(alpha, task_mode, history, output, target) → None

class colda.algorithm.train_stage.api.MakeTrain

Bases: BaseAlgorithm

Train the model

Methods

make_train

classmethod make_train(dataset_path: str, data_idx: str, skip_header: int, task_mode: Literal['classification', 'regression'], model_name: Literal['linear', 'decision_tree', 'svm', 'gradient_boosting', 'mlp'], cur_round_residual: Any, role: Literal['sponsor', 'assistor'], matched_identifier: Any | None = None) → tuple[Any, numpy.ndarray]

class colda.algorithm.train_stage.api.MakeTrainLocal

Bases: BaseAlgorithm

Sponsor train model locally(only use its dataset) and get its performance

Methods

make_train_local

classmethod make_train_local(root: str, self_id: str, task_id: str, dataset_path: str, data_idx: str, target_idx: str, skip_header: str, task_mode: Literal['classification', 'regression'], model_name: Literal['linear', 'decision_tree', 'svm', 'gradient_boosting', 'mlp'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC']) → None

colda.algorithm.train_stage.make_residual module

class colda.algorithm.train_stage.make_residual.MakeResidual

Bases: BaseAlgorithm

Calculate Residual. Residual is used as target for the following training step.

Attributes

None

Methods

make_residual

classmethod compute_residual(task_mode: Literal['classification', 'regression'], output: ndarray, target: ndarray) → ndarray

Compute residual based on task mode, output and target

Parameters

task_mode : str target : list[str]

Returns

np.ndarray

classmethod make_init(task_mode: Literal['classification', 'regression'], target: list[str]) → ndarray

Create initial residual at round 1

Parameters

task_mode : str target : list[str]

Returns

np.ndarray

classmethod make_residual(round: str, dataset_path: str, target_idx: str, skip_header: str, task_mode: Literal['classification', 'regression'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC'], sponsor_matched_identifers: dict[str, Any], last_round_result: None) → ndarray

classmethod make_residual(round: str, dataset_path: str, target_idx: str, skip_header: str, task_mode: Literal['classification', 'regression'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC'], sponsor_matched_identifers: dict[str, Any], last_round_result: Any) → ndarray

Calculate Residual.

Parameters

round : str dataset_path : str target_idx : str skip_header : str task_mode : str metric_name : str sponsor_matched_identifers : dict[str, Any] last_round_result : Union[Any, None]=None

Returns

tuple[np.ndarray[Any], dict[str, Any]]

colda.algorithm.train_stage.make_result module

class colda.algorithm.train_stage.make_result.MakeResult

Bases: BaseAlgorithm

Combine the sponsor’s trained model and assistors’ trained models to a better sponsor model

Methods

make_result

classmethod make_result(user_id: str, train_id: str, rounds, dataset_path, target_idx, skip_header, task_mode: Literal['classification', 'regression'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC'], sponsor_trained_cooperative_model_output, assistor_trained_cooperative_model_outputs, sponsor_matched_identifers, last_round_result) → None

classmethod result_func(alpha, task_mode, history, output, target) → None

colda.algorithm.train_stage.make_train module

class colda.algorithm.train_stage.make_train.MakeTrain

Bases: BaseAlgorithm

Train the model

Methods

make_train

classmethod make_train(dataset_path: str, data_idx: str, skip_header: int, task_mode: Literal['classification', 'regression'], model_name: Literal['linear', 'decision_tree', 'svm', 'gradient_boosting', 'mlp'], cur_round_residual: Any, role: Literal['sponsor', 'assistor'], matched_identifier: Any | None = None) → tuple[Any, numpy.ndarray]

colda.algorithm.train_stage.make_train_local module

class colda.algorithm.train_stage.make_train_local.MakeTrainLocal

Bases: BaseAlgorithm

Sponsor train model locally(only use its dataset) and get its performance

Methods

make_train_local

classmethod make_train_local(root: str, self_id: str, task_id: str, dataset_path: str, data_idx: str, target_idx: str, skip_header: str, task_mode: Literal['classification', 'regression'], model_name: Literal['linear', 'decision_tree', 'svm', 'gradient_boosting', 'mlp'], metric_name: Literal['MAD', 'RMSE', 'R2', 'Accuracy', 'F1', 'AUCROC']) → None

Module contents