coco_pipe.decoding.configs

Comprehensive Pydantic models for strict validation of Decoding/ML experiments. These models ensure that all parameters are scientifically sound before any computation begins.

Attributes

AtomicEstimator
EstimatorConfigType
ClassicalModelType
ModelConfigType

Classes

BaseEstimatorConfig	Base configuration for any estimator.
ClassicalEstimatorConfig	Base for scikit-learn compatible classical estimators.
LinearMixin	Common parameters for linear models.
RegularizedLinearMixin	Parameters for regularized linear models.
TreeMixin	Common parameters for Tree-based models.
SupportVectorMixin	Common parameters for Support Vector Machines.
SGDMixin	Stochastic Gradient Descent parameters.
MLPMixin	Common parameters for Multi-layer Perceptron models.
GradientBoostingMixin	Common parameters for Gradient Boosting models.
LogisticRegressionConfig	Configuration for sklearn.linear_model.LogisticRegression.
RandomForestClassifierConfig	Configuration for sklearn.ensemble.RandomForestClassifier.
SVCConfig	Configuration for sklearn.svm.SVC.
LinearSVCConfig	Configuration for sklearn.svm.LinearSVC.
KNeighborsClassifierConfig	Configuration for sklearn.neighbors.KNeighborsClassifier.
GradientBoostingClassifierConfig	Configuration for sklearn.ensemble.GradientBoostingClassifier.
HistGradientBoostingClassifierConfig	Configuration for sklearn.ensemble.HistGradientBoostingClassifier.
SGDClassifierConfig	Configuration for sklearn.linear_model.SGDClassifier.
MLPClassifierConfig	Configuration for sklearn.neural_network.MLPClassifier.
GaussianNBConfig	Configuration for sklearn.naive_bayes.GaussianNB.
LDAConfig	Configuration for sklearn.discriminant_analysis.LinearDiscriminantAnalysis.
AdaBoostClassifierConfig	Configuration for sklearn.ensemble.AdaBoostClassifier.
DummyClassifierConfig	Configuration for sklearn.dummy.DummyClassifier.
LPFTConfig	Configuration for Linear-Probe Fine-Tuning (LP-FT).
SkorchClassifierConfig	Configuration for generic PyTorch wrappers via Skorch.
SlidingEstimatorConfig	Configuration for MNE-style SlidingEstimator.
GeneralizingEstimatorConfig	Configuration for MNE-style GeneralizingEstimator.
LinearRegressionConfig	Configuration for sklearn.linear_model.LinearRegression.
RidgeConfig	Configuration for sklearn.linear_model.Ridge.
LassoConfig	Configuration for sklearn.linear_model.Lasso.
ElasticNetConfig	Configuration for sklearn.linear_model.ElasticNet.
RandomForestRegressorConfig	Configuration for sklearn.ensemble.RandomForestRegressor.
SVRConfig	Configuration for sklearn.svm.SVR.
GradientBoostingRegressorConfig	Configuration for sklearn.ensemble.GradientBoostingRegressor.
SGDRegressorConfig	Configuration for sklearn.linear_model.SGDRegressor.
MLPRegressorConfig	Configuration for sklearn.neural_network.MLPRegressor.
DummyRegressorConfig	Configuration for sklearn.dummy.DummyRegressor.
DecisionTreeRegressorConfig	Configuration for sklearn.tree.DecisionTreeRegressor.
KNeighborsRegressorConfig	Configuration for sklearn.neighbors.KNeighborsRegressor.
ExtraTreesRegressorConfig	Configuration for sklearn.ensemble.ExtraTreesRegressor.
HistGradientBoostingRegressorConfig	Configuration for sklearn.ensemble.HistGradientBoostingRegressor.
AdaBoostRegressorConfig	Configuration for sklearn.ensemble.AdaBoostRegressor.
BayesianRidgeConfig	Configuration for sklearn.linear_model.BayesianRidge.
ARDRegressionConfig	Configuration for sklearn.linear_model.ARDRegression.
ClassicalModelConfig	Configuration for standard scikit-learn estimators.
FoundationEmbeddingModelConfig	Configuration for pretrained feature extraction backbones.
LoRAConfig	Low-Rank Adaptation (LoRA) configuration.
QuantizationConfig	Model quantization settings.
DeviceConfig	Compute device and precision policy.
CheckpointConfig	Neural training checkpoint policy.
TrainerConfig	Neural training loop configuration.
FrozenBackboneDecoderConfig	Config for a frozen backbone followed by a classical decoding head.
NeuralFineTuneConfig	Configuration for end-to-end neural fine-tuning.
TemporalDecoderConfig	Config for MNE-style temporal meta-estimators.
CVConfig	Cross-validation configuration settings.
TuningConfig	Hyperparameter Tuning Configuration.
FeatureSelectionConfig	Feature selection settings.
ReducerConfig	Fold-local dimensionality reduction for classical decoding.
CalibrationConfig	Probability calibration settings for classification estimators.
ConfidenceIntervalConfig	Analytical confidence interval settings.
ChanceAssessmentConfig	Null-hypothesis (chance level) assessment settings.
StatisticalAssessmentConfig	Settings for finite-sample statistical inference and uncertainty estimation.

Module Contents

class coco_pipe.decoding.configs.BaseEstimatorConfig(/, **data)

Bases: pydantic.BaseModel

Base configuration for any estimator.

Parameters:

data (Any)

class coco_pipe.decoding.configs.ClassicalEstimatorConfig(/, **data)

Bases: BaseEstimatorConfig

Base for scikit-learn compatible classical estimators.

Parameters:

data (Any)

kind: Literal['classical'] = 'classical'

class coco_pipe.decoding.configs.LinearMixin(/, **data)

Bases: pydantic.BaseModel

Common parameters for linear models.

Parameters:

data (Any)

fit_intercept: bool = True

copy_X: bool = True

n_jobs: int | None = None

class coco_pipe.decoding.configs.RegularizedLinearMixin(/, **data)

Bases: pydantic.BaseModel

Parameters for regularized linear models.

Parameters:

data (Any)

fit_intercept: bool = True

copy_X: bool = True

tol: float = 0.001

max_iter: int | None = None

positive: bool = False

random_state: int | None = None

class coco_pipe.decoding.configs.TreeMixin(/, **data)

Bases: pydantic.BaseModel

Common parameters for Tree-based models.

Parameters:

data (Any)

n_estimators: int = None

max_depth: int | None = None

min_samples_split: int | float = 2

min_samples_leaf: int | float = 1

min_weight_fraction_leaf: float = 0.0

max_features: str | int | float | None = 'sqrt'

max_leaf_nodes: int | None = None

min_impurity_decrease: float = 0.0

ccp_alpha: float = 0.0

n_jobs: int | None = None

random_state: int | None = None

verbose: int = 0

warm_start: bool = False

class coco_pipe.decoding.configs.SupportVectorMixin(/, **data)

Bases: pydantic.BaseModel

Common parameters for Support Vector Machines.

Parameters:

data (Any)

C: float = None

kernel: Literal['linear', 'poly', 'rbf', 'sigmoid', 'precomputed'] = 'rbf'

degree: int = 3

gamma: str | float = 'scale'

coef0: float = 0.0

tol: float = 0.001

verbose: bool = False

max_iter: int = -1

shrinking: bool = True

cache_size: float = 200

class coco_pipe.decoding.configs.SGDMixin(/, **data)

Bases: pydantic.BaseModel

Stochastic Gradient Descent parameters.

Parameters:

data (Any)

loss: str = 'hinge'

penalty: Literal['l2', 'l1', 'elasticnet'] | None = 'l2'

alpha: float = 0.0001

l1_ratio: float = 0.15

fit_intercept: bool = True

max_iter: int = 1000

tol: float = 0.001

shuffle: bool = True

verbose: int = 0

epsilon: float = 0.1

learning_rate: str = 'optimal'

eta0: float = 0.01

power_t: float = 0.5

early_stopping: bool = False

class coco_pipe.decoding.configs.MLPMixin(/, **data)

Bases: pydantic.BaseModel

Common parameters for Multi-layer Perceptron models.

Parameters:

data (Any)

hidden_layer_sizes: tuple = (100,)

activation: Literal['identity', 'logistic', 'tanh', 'relu'] = 'relu'

solver: Literal['lbfgs', 'sgd', 'adam'] = 'adam'

alpha: float = 0.0001

batch_size: int | str = 'auto'

learning_rate: Literal['constant', 'invscaling', 'adaptive'] = 'constant'

learning_rate_init: float = 0.001

power_t: float = 0.5

max_iter: int = 200

shuffle: bool = True

tol: float = 0.0001

verbose: bool = False

warm_start: bool = False

momentum: float = 0.9

nesterovs_momentum: bool = True

early_stopping: bool = False

validation_fraction: float = 0.1

beta_1: float = 0.9

beta_2: float = 0.999

epsilon: float = 1e-08

n_iter_no_change: int = 10

max_fun: int = 15000

random_state: int | None = None

class coco_pipe.decoding.configs.GradientBoostingMixin(/, **data)

Bases: pydantic.BaseModel

Common parameters for Gradient Boosting models.

Parameters:

data (Any)

learning_rate: float = 0.1

n_estimators: int = 100

subsample: float = 1.0

criterion: Literal['friedman_mse', 'squared_error'] = 'friedman_mse'

min_samples_split: int | float = 2

min_samples_leaf: int | float = 1

min_weight_fraction_leaf: float = 0.0

max_depth: int = 3

min_impurity_decrease: float = 0.0

init: str | None = None

max_features: str | int | float | None = None

verbose: int = 0

max_leaf_nodes: int | None = None

warm_start: bool = False

validation_fraction: float = 0.1

n_iter_no_change: int = 5

tol: float = 0.0001

ccp_alpha: float = 0.0

random_state: int | None = None

class coco_pipe.decoding.configs.LogisticRegressionConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.linear_model.LogisticRegression.

Parameters:

data (Any)

method: Literal['LogisticRegression'] = 'LogisticRegression'

penalty: Literal['l1', 'l2', 'elasticnet', None] = 'l2'

dual: bool = False

tol: float = 0.0001

C: float = None

fit_intercept: bool = True

intercept_scaling: float = 1.0

class_weight: dict | str | None = None

solver: Literal['newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga'] = 'lbfgs'

max_iter: int = 100

verbose: int = 0

warm_start: bool = False

n_jobs: int | None = None

l1_ratio: float | None = None

random_state: int | None = None

class coco_pipe.decoding.configs.RandomForestClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig, TreeMixin

Configuration for sklearn.ensemble.RandomForestClassifier.

Parameters:

data (Any)

method: Literal['RandomForestClassifier'] = 'RandomForestClassifier'

criterion: Literal['gini', 'entropy', 'log_loss'] = 'gini'

bootstrap: bool = True

oob_score: bool = False

class_weight: str | dict | list | None = None

max_samples: int | float | None = None

class coco_pipe.decoding.configs.SVCConfig(/, **data)

Bases: ClassicalEstimatorConfig, SupportVectorMixin

Configuration for sklearn.svm.SVC.

Parameters:

data (Any)

method: Literal['SVC'] = 'SVC'

probability: bool = True

class_weight: dict | str | None = None

decision_function_shape: Literal['ovo', 'ovr'] = 'ovr'

break_ties: bool = False

random_state: int | None = None

class coco_pipe.decoding.configs.LinearSVCConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.svm.LinearSVC.

Parameters:

data (Any)

method: Literal['LinearSVC'] = 'LinearSVC'

penalty: Literal['l1', 'l2'] = 'l2'

loss: Literal['hinge', 'squared_hinge'] = 'squared_hinge'

dual: bool | Literal['auto'] = 'auto'

tol: float = 0.0001

C: float = None

multi_class: Literal['ovr', 'crammer_singer'] = 'ovr'

fit_intercept: bool = True

intercept_scaling: float = 1.0

class_weight: dict | str | None = None

verbose: int = 0

max_iter: int = 1000

random_state: int | None = None

class coco_pipe.decoding.configs.KNeighborsClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.neighbors.KNeighborsClassifier.

Parameters:

data (Any)

method: Literal['KNeighborsClassifier'] = 'KNeighborsClassifier'

n_neighbors: int = None

weights: Literal['uniform', 'distance'] = 'uniform'

algorithm: Literal['auto', 'ball_tree', 'kd_tree', 'brute'] = 'auto'

leaf_size: int = 30

p: int = 2

metric: str = 'minkowski'

metric_params: dict | None = None

n_jobs: int | None = None

class coco_pipe.decoding.configs.GradientBoostingClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig, GradientBoostingMixin

Configuration for sklearn.ensemble.GradientBoostingClassifier.

Parameters:

data (Any)

method: Literal['GradientBoostingClassifier'] = 'GradientBoostingClassifier'

loss: Literal['log_loss', 'exponential'] = 'log_loss'

class coco_pipe.decoding.configs.HistGradientBoostingClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.ensemble.HistGradientBoostingClassifier.

Parameters:

data (Any)

method: Literal['HistGradientBoostingClassifier'] = 'HistGradientBoostingClassifier'

learning_rate: float = 0.1

max_iter: int = 100

max_leaf_nodes: int = 31

max_depth: int | None = None

min_samples_leaf: int = 20

l2_regularization: float = 0.0

max_bins: int = 255

categorical_features: list[int] | list[str] | list[bool] | None = None

monotonic_cst: Any | None = None

interaction_cst: Any | None = None

warm_start: bool = False

early_stopping: bool | Literal['auto'] = 'auto'

scoring: str | None = 'loss'

validation_fraction: float = 0.1

n_iter_no_change: int = 10

tol: float = 1e-07

verbose: int = 0

random_state: int | None = None

class coco_pipe.decoding.configs.SGDClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig, SGDMixin

Configuration for sklearn.linear_model.SGDClassifier.

Parameters:

data (Any)

method: Literal['SGDClassifier'] = 'SGDClassifier'

class_weight: dict | str | None = None

class coco_pipe.decoding.configs.MLPClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig, MLPMixin

Configuration for sklearn.neural_network.MLPClassifier.

Parameters:

data (Any)

method: Literal['MLPClassifier'] = 'MLPClassifier'

class coco_pipe.decoding.configs.GaussianNBConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.naive_bayes.GaussianNB.

Parameters:

data (Any)

method: Literal['GaussianNB'] = 'GaussianNB'

priors: list[float] | None = None

var_smoothing: float = 1e-09

class coco_pipe.decoding.configs.LDAConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.discriminant_analysis.LinearDiscriminantAnalysis.

Parameters:

data (Any)

method: Literal['LinearDiscriminantAnalysis'] = 'LinearDiscriminantAnalysis'

solver: Literal['svd', 'lsqr', 'eigen'] = 'svd'

shrinkage: str | float | None = None

priors: list[float] | None = None

n_components: int | None = None

store_covariance: bool = False

tol: float = 0.0001

class coco_pipe.decoding.configs.AdaBoostClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.ensemble.AdaBoostClassifier.

Parameters:

data (Any)

method: Literal['AdaBoostClassifier'] = 'AdaBoostClassifier'

n_estimators: int = 50

learning_rate: float = 1.0

random_state: int | None = None

class coco_pipe.decoding.configs.DummyClassifierConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.dummy.DummyClassifier.

Parameters:

data (Any)

method: Literal['DummyClassifier'] = 'DummyClassifier'

strategy: Literal['stratified', 'most_frequent', 'prior', 'uniform'] = 'prior'

constant: Any | None = None

random_state: int | None = None

class coco_pipe.decoding.configs.LPFTConfig(/, **data)

Bases: BaseEstimatorConfig

Configuration for Linear-Probe Fine-Tuning (LP-FT). Reference: Kumar et al. (2022).

Parameters:

data (Any)

method: Literal['LPFTClassifier'] = 'LPFTClassifier'

backbone_name: str = 'gpt2'

lp_lr: float = 0.001

lp_epochs: int = 10

ft_lr: float = 1e-05

ft_epochs: int = 5

batch_size: int = 32

max_length: int = 128

device: str = 'cpu'

class coco_pipe.decoding.configs.SkorchClassifierConfig(/, **data)

Bases: BaseEstimatorConfig

Configuration for generic PyTorch wrappers via Skorch.

Parameters:

data (Any)

method: Literal['SkorchClassifier'] = 'SkorchClassifier'

module_name: str

max_epochs: int = 10

lr: float = 0.01

batch_size: int = 64

optimizer: str = 'Adam'

device: str = 'cpu'

class coco_pipe.decoding.configs.SlidingEstimatorConfig(/, **data)

Bases: BaseEstimatorConfig

Configuration for MNE-style SlidingEstimator. Fits a separate estimator for each time point.

Parameters:

data (Any)

method: Literal['SlidingEstimator'] = 'SlidingEstimator'

base_estimator: EstimatorConfigType

scoring: str | collections.abc.Callable | None = None

n_jobs: int | None = 1

position: float | None = 0

allow_2d: bool = False

verbose: bool | str | int | None = None

class coco_pipe.decoding.configs.GeneralizingEstimatorConfig(/, **data)

Bases: BaseEstimatorConfig

Configuration for MNE-style GeneralizingEstimator. Fits an estimator on each time point and tests on all other time points.

Parameters:

data (Any)

method: Literal['GeneralizingEstimator'] = 'GeneralizingEstimator'

base_estimator: EstimatorConfigType

scoring: str | collections.abc.Callable | None = None

n_jobs: int | None = 1

position: float | None = 0

allow_2d: bool = False

verbose: bool | str | int | None = None

class coco_pipe.decoding.configs.LinearRegressionConfig(/, **data)

Bases: ClassicalEstimatorConfig, LinearMixin

Configuration for sklearn.linear_model.LinearRegression.

Parameters:

data (Any)

method: Literal['LinearRegression'] = 'LinearRegression'

positive: bool = False

class coco_pipe.decoding.configs.RidgeConfig(/, **data)

Bases: ClassicalEstimatorConfig, RegularizedLinearMixin

Configuration for sklearn.linear_model.Ridge.

Parameters:

data (Any)

method: Literal['Ridge'] = 'Ridge'

alpha: float = 1.0

fit_intercept: bool = True

copy_X: bool = True

solver: str = 'auto'

class coco_pipe.decoding.configs.LassoConfig(/, **data)

Bases: ClassicalEstimatorConfig, RegularizedLinearMixin

Configuration for sklearn.linear_model.Lasso.

Parameters:

data (Any)

method: Literal['Lasso'] = 'Lasso'

alpha: float = 1.0

precompute: bool | list = False

fit_intercept: bool = True

copy_X: bool = True

selection: Literal['cyclic', 'random'] = 'cyclic'

warm_start: bool = False

class coco_pipe.decoding.configs.ElasticNetConfig(/, **data)

Bases: ClassicalEstimatorConfig, RegularizedLinearMixin

Configuration for sklearn.linear_model.ElasticNet.

Parameters:

data (Any)

method: Literal['ElasticNet'] = 'ElasticNet'

alpha: float = 1.0

l1_ratio: float = 0.5

precompute: bool | list = False

fit_intercept: bool = True

copy_X: bool = True

selection: Literal['cyclic', 'random'] = 'cyclic'

warm_start: bool = False

class coco_pipe.decoding.configs.RandomForestRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig, TreeMixin

Configuration for sklearn.ensemble.RandomForestRegressor.

Parameters:

data (Any)

method: Literal['RandomForestRegressor'] = 'RandomForestRegressor'

criterion: Literal['squared_error', 'absolute_error', 'friedman_mse', 'poisson'] = 'squared_error'

bootstrap: bool = True

oob_score: bool = False

max_samples: int | float | None = None

class coco_pipe.decoding.configs.SVRConfig(/, **data)

Bases: ClassicalEstimatorConfig, SupportVectorMixin

Configuration for sklearn.svm.SVR.

Parameters:

data (Any)

method: Literal['SVR'] = 'SVR'

epsilon: float = 0.1

class coco_pipe.decoding.configs.GradientBoostingRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig, GradientBoostingMixin

Configuration for sklearn.ensemble.GradientBoostingRegressor.

Parameters:

data (Any)

method: Literal['GradientBoostingRegressor'] = 'GradientBoostingRegressor'

loss: Literal['squared_error', 'absolute_error', 'huber', 'quantile'] = 'squared_error'

alpha: float = 0.9

class coco_pipe.decoding.configs.SGDRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig, SGDMixin

Configuration for sklearn.linear_model.SGDRegressor.

Parameters:

data (Any)

method: Literal['SGDRegressor'] = 'SGDRegressor'

loss: str = 'squared_error'

class coco_pipe.decoding.configs.MLPRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig, MLPMixin

Configuration for sklearn.neural_network.MLPRegressor.

Parameters:

data (Any)

method: Literal['MLPRegressor'] = 'MLPRegressor'

class coco_pipe.decoding.configs.DummyRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.dummy.DummyRegressor.

Parameters:

data (Any)

method: Literal['DummyRegressor'] = 'DummyRegressor'

strategy: Literal['mean', 'median', 'quantile', 'constant'] = 'mean'

constant: int | float | list | None = None

quantile: float | None = None

class coco_pipe.decoding.configs.DecisionTreeRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.tree.DecisionTreeRegressor.

Parameters:

data (Any)

method: Literal['DecisionTreeRegressor'] = 'DecisionTreeRegressor'

criterion: Literal['squared_error', 'friedman_mse', 'absolute_error', 'poisson'] = 'squared_error'

splitter: Literal['best', 'random'] = 'best'

max_depth: int | None = None

min_samples_split: int | float = 2

min_samples_leaf: int | float = 1

min_weight_fraction_leaf: float = 0.0

max_features: str | int | float | None = None

random_state: int | None = None

max_leaf_nodes: int | None = None

min_impurity_decrease: float = 0.0

ccp_alpha: float = 0.0

class coco_pipe.decoding.configs.KNeighborsRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.neighbors.KNeighborsRegressor.

Parameters:

data (Any)

method: Literal['KNeighborsRegressor'] = 'KNeighborsRegressor'

n_neighbors: int = None

weights: Literal['uniform', 'distance'] = 'uniform'

algorithm: Literal['auto', 'ball_tree', 'kd_tree', 'brute'] = 'auto'

leaf_size: int = 30

p: int = 2

metric: str = 'minkowski'

metric_params: dict | None = None

n_jobs: int | None = None

class coco_pipe.decoding.configs.ExtraTreesRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig, TreeMixin

Configuration for sklearn.ensemble.ExtraTreesRegressor.

Parameters:

data (Any)

method: Literal['ExtraTreesRegressor'] = 'ExtraTreesRegressor'

bootstrap: bool = False

oob_score: bool = False

max_samples: int | float | None = None

class coco_pipe.decoding.configs.HistGradientBoostingRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.ensemble.HistGradientBoostingRegressor.

Parameters:

data (Any)

method: Literal['HistGradientBoostingRegressor'] = 'HistGradientBoostingRegressor'

loss: Literal['squared_error', 'absolute_error', 'poisson', 'quantile'] = 'squared_error'

learning_rate: float = 0.1

max_iter: int = 100

max_leaf_nodes: int = 31

max_depth: int | None = None

min_samples_leaf: int = 20

l2_regularization: float = 0.0

max_bins: int = 255

categorical_features: list[int] | list[str] | list[bool] | None = None

monotonic_cst: Any | None = None

interaction_cst: Any | None = None

warm_start: bool = False

early_stopping: bool | Literal['auto'] = 'auto'

scoring: str | None = 'loss'

validation_fraction: float = 0.1

n_iter_no_change: int = 10

tol: float = 1e-07

verbose: int = 0

random_state: int | None = None

class coco_pipe.decoding.configs.AdaBoostRegressorConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.ensemble.AdaBoostRegressor.

Parameters:

data (Any)

method: Literal['AdaBoostRegressor'] = 'AdaBoostRegressor'

n_estimators: int = 50

learning_rate: float = 1.0

loss: Literal['linear', 'square', 'exponential'] = 'linear'

random_state: int | None = None

class coco_pipe.decoding.configs.BayesianRidgeConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.linear_model.BayesianRidge.

Parameters:

data (Any)

method: Literal['BayesianRidge'] = 'BayesianRidge'

max_iter: int = 300

tol: float = 0.001

alpha_1: float = 1e-06

alpha_2: float = 1e-06

lambda_1: float = 1e-06

lambda_2: float = 1e-06

alpha_init: float | None = None

lambda_init: float | None = None

compute_score: bool = False

fit_intercept: bool = True

copy_X: bool = True

verbose: bool = False

class coco_pipe.decoding.configs.ARDRegressionConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for sklearn.linear_model.ARDRegression.

Parameters:

data (Any)

method: Literal['ARDRegression'] = 'ARDRegression'

max_iter: int = 300

tol: float = 0.001

alpha_1: float = 1e-06

alpha_2: float = 1e-06

lambda_1: float = 1e-06

lambda_2: float = 1e-06

compute_score: bool = False

threshold_lambda: float = 10000.0

fit_intercept: bool = True

copy_X: bool = True

verbose: bool = False

class coco_pipe.decoding.configs.ClassicalModelConfig(/, **data)

Bases: ClassicalEstimatorConfig

Configuration for standard scikit-learn estimators.

Parameters:

data (Any)

method: Literal['ClassicalModel'] = 'ClassicalModel'

estimator: str

params: dict[str, Any] = None

input_kind: Literal['tabular', 'embeddings'] = 'tabular'

class coco_pipe.decoding.configs.FoundationEmbeddingModelConfig(/, **data)

Bases: BaseEstimatorConfig

Configuration for pretrained feature extraction backbones.

Parameters:

data (Any)

kind: Literal['foundation_embedding'] = 'foundation_embedding'

backend: str = 'auto'

n_outputs: int | None = None

train_mode: Literal['frozen', 'full', 'lora', 'qlora'] = 'frozen'

pooling: Literal['mean', 'flatten'] = 'mean'

normalize_embeddings: bool = True

cache_embeddings: bool = True

sfreq: float | None = None

ch_names: list[str] | None = None

backend_kwargs: dict[str, Any] = None

class coco_pipe.decoding.configs.LoRAConfig(/, **data)

Bases: pydantic.BaseModel

Low-Rank Adaptation (LoRA) configuration.

Parameters:

data (Any)

r: int = None

alpha: int = None

dropout: float = None

target_modules: str | list[str] = 'all-linear'

class coco_pipe.decoding.configs.QuantizationConfig(/, **data)

Bases: pydantic.BaseModel

Model quantization settings.

Parameters:

data (Any)

enabled: bool = False

load_in_4bit: bool = True

quant_type: Literal['nf4', 'fp4'] = 'nf4'

compute_dtype: Literal['bf16', 'fp16', 'fp32'] = 'bf16'

class coco_pipe.decoding.configs.DeviceConfig(/, **data)

Bases: pydantic.BaseModel

Compute device and precision policy.

Parameters:

data (Any)

device: Literal['auto', 'cpu', 'cuda', 'mps'] = 'auto'

precision: Literal['fp32', 'fp16', 'bf16'] = 'fp32'

class coco_pipe.decoding.configs.CheckpointConfig(/, **data)

Bases: pydantic.BaseModel

Neural training checkpoint policy.

Parameters:

data (Any)

save: Literal['none', 'best', 'last', 'all'] = 'best'

monitor: str = 'val_loss'

output_dir: pathlib.Path | None = None

class coco_pipe.decoding.configs.TrainerConfig(/, **data)

Bases: pydantic.BaseModel

Neural training loop configuration.

Parameters:

data (Any)

max_epochs: int = None

early_stopping_patience: int | None = None

batch_size: int = None

validation_fraction: float = None

class coco_pipe.decoding.configs.FrozenBackboneDecoderConfig(/, **data)

Bases: BaseEstimatorConfig

Config for a frozen backbone followed by a classical decoding head.

Parameters:

data (Any)

kind: Literal['frozen_backbone'] = 'frozen_backbone'

backbone: FoundationEmbeddingModelConfig

head: ClassicalModelConfig

class coco_pipe.decoding.configs.NeuralFineTuneConfig(/, **data)

Bases: BaseEstimatorConfig

Configuration for end-to-end neural fine-tuning.

Parameters:

data (Any)

kind: Literal['neural_finetune'] = 'neural_finetune'

backend: str = 'auto'

n_outputs: int | None = None

input_kind: Literal['temporal', 'epoched', 'tokens'] = 'epoched'

train_mode: Literal['full', 'frozen', 'linear_probe', 'lora', 'qlora'] = 'full'

optimizer: dict[str, Any] = None

trainer: TrainerConfig = None

device: DeviceConfig = None

checkpoints: CheckpointConfig = None

lora: LoRAConfig | None = None

quantization: QuantizationConfig | None = None

sfreq: float | None = None

ch_names: list[str] | None = None

backend_kwargs: dict[str, Any] = None

class_weight: str | dict[Any, float] | None = 'balanced'

class coco_pipe.decoding.configs.TemporalDecoderConfig(/, **data)

Bases: BaseEstimatorConfig

Config for MNE-style temporal meta-estimators.

Parameters:

data (Any)

kind: Literal['temporal'] = 'temporal'

wrapper: Literal['sliding', 'generalizing'] = 'sliding'

base: ClassicalModelConfig

scoring: str | collections.abc.Callable | None = None

n_jobs: int | None = 1

position: float | None = 0

allow_2d: bool = False

verbose: bool | str | int | None = None

coco_pipe.decoding.configs.AtomicEstimator

coco_pipe.decoding.configs.EstimatorConfigType

coco_pipe.decoding.configs.ClassicalModelType

class coco_pipe.decoding.configs.CVConfig(/, **data)

Bases: pydantic.BaseModel

Cross-validation configuration settings.

Parameters:

• strategy (str, default="stratified") – The splitting strategy. Note that ‘stratified’ strategies require classification tasks.

• n_splits (int, default=5) – Number of folds. Must be at least 2.

• shuffle (bool, default=True) – Whether to shuffle data before splitting.

• random_state (int, default=42) – Random seed for the splitter.

• test_size (float, default=0.2) – The proportion of the dataset to include in the test split for strategy=’split’.

• stratify (bool, default=False) – Whether to use stratified sampling for strategy=’split’.

• group_key (str, optional) – The column name in sample_metadata to use for group-aware strategies.

• data (Any)

strategy: Literal['stratified', 'kfold', 'group_kfold', 'stratified_group_kfold', 'leave_p_out', 'leave_one_group_out', 'timeseries', 'split', 'group_shuffle_split'] = 'stratified'

n_splits: int = None

shuffle: bool = True

random_state: int = 42

test_size: float = None

stratify: bool = None

group_key: str | None = None

auto_reduce_n_splits: bool = None

class coco_pipe.decoding.configs.TuningConfig(/, **data)

Bases: pydantic.BaseModel

Hyperparameter Tuning Configuration.

Parameters:

data (Any)

enabled: bool = False

search_type: Literal['grid', 'random'] = 'grid'

n_iter: int = None

scoring: str | None = None

n_jobs: int = -1

random_state: int | None = None

cv: CVConfig | None = None

allow_nongroup_inner_cv: bool = None

class coco_pipe.decoding.configs.FeatureSelectionConfig(/, **data)

Bases: pydantic.BaseModel

Feature selection settings.

Parameters:

data (Any)

enabled: bool = False

method: Literal['k_best', 'sfs'] = 'sfs'

n_features: int | None = None

direction: Literal['forward', 'backward'] = 'forward'

tol: float | None = None

cv: CVConfig | None = None

scoring: str | None = None

allow_nongroup_inner_cv: bool = None

class coco_pipe.decoding.configs.ReducerConfig(/, **data)

Bases: pydantic.BaseModel

Fold-local dimensionality reduction for classical decoding.

Parameters:

data (Any)

enabled: bool = False

method: Literal['pca'] = 'pca'

n_components: int | float | None = None

whiten: bool = False

svd_solver: Literal['auto', 'full', 'covariance_eigh', 'arpack', 'randomized'] = 'auto'

random_state: int | None = 42

class coco_pipe.decoding.configs.CalibrationConfig(/, **data)

Bases: pydantic.BaseModel

Probability calibration settings for classification estimators.

Parameters:

data (Any)

enabled: bool = False

method: Literal['sigmoid', 'isotonic'] = 'sigmoid'

cv: CVConfig | None = None

n_jobs: int | None = None

allow_nongroup_inner_cv: bool = None

class coco_pipe.decoding.configs.ConfidenceIntervalConfig(/, **data)

Bases: pydantic.BaseModel

Analytical confidence interval settings.

Parameters:

data (Any)

alpha: float = None

method: Literal['wilson', 'clopper_pearson'] = 'wilson'

class coco_pipe.decoding.configs.ChanceAssessmentConfig(/, **data)

Bases: pydantic.BaseModel

Null-hypothesis (chance level) assessment settings.

Parameters:

data (Any)

method: Literal['permutation', 'binomial', 'auto'] = 'permutation'

n_permutations: int = None

p0: float | Literal['auto'] | None = None

temporal_correction: Literal['max_stat', 'fdr_bh', 'none'] = None

store_null_distribution: bool = False

class coco_pipe.decoding.configs.StatisticalAssessmentConfig(/, **data)

Bases: pydantic.BaseModel

Settings for finite-sample statistical inference and uncertainty estimation.

Parameters:

data (Any)

enabled: bool = False

random_state: int | None = 42

metrics: list[str] | None = None

chance: ChanceAssessmentConfig = None

confidence_intervals: ConfidenceIntervalConfig = None

unit_of_inference: Literal['sample', 'group_mean', 'group_majority', 'custom'] | None = None

custom_unit_column: str | None = None

custom_aggregation: Literal['mean', 'majority'] = 'mean'

coco_pipe.decoding.configs.ModelConfigType