Deep learning for NeuroImaging in Python.

Note

This page is a reference documentation. It only explains the class signature, and not how to use it. Please refer to the gallery for the big picture.

class nidl.callbacks.model_probing.LogisticRegressionCVCallback(train_dataloader: DataLoader, test_dataloader: DataLoader, probe_name: str | None = None, Cs: int | list[float] = 5, cv: int = 5, max_iter: int = 100, n_jobs: int | None = None, scoring: str = 'balanced_accuracy', linear_solver: str = 'lbfgs', **kwargs)[source]¶

Bases: ModelProbing

Performs logistic regression on top of an embedding model.

Concretely this callback:

Embeds the input data through the torch model.
Performs n-fold CV to find the best L2 regularization strength.
Logs the main classification metrics for each class and averaged across classes (weighted by class support and unweighted):
- precision
- recall
- f1-score
- support
- accuracy (global)

Please check this User Guide for more details on the classification metrics reported.

Parameters:

train_dataloader : torch.utils.data.DataLoader

Training dataloader yielding batches in the form (X, y) for further embedding and training of the logistic regression probe.

test_dataloader : torch.utils.data.DataLoader

Test dataloader yielding batches in the form (X, y) for further embedding and test of the logistic regression probe.

probe_name : str or None, default=None

Name of the probe displayed when logging the results. It will appear as <probe_name>/<metric_name> for each metric. If None, only <metric_name> is displayed.

Cs : int or list of floats, default=10

Each of the values in Cs describes the inverse of regularization strength. If Cs is as an int, then a grid of Cs values are chosen in a logarithmic scale between 1e-4 and 1e4. Like in support vector machines, smaller values specify stronger regularization.

cv : int or cross-validation generator, default=5

How many folds to use for cross-validating the C regularization strenght in the LogisticRegression.

max_iter : int, default=100

Maximum number of iterations taken for the solver to converge.

n_jobs : int or None, default=None

Number of jobs to run in parallel. None means 1 unless in a joblib.parallel_backend context. -1 means using all processors.

scoring : str in {“accuracy”, “balanced_accuracy”, “f1”, …},

default=”balanced_accuracy” Which scoring function to use to cross-validate the C hyper-parameter. For a complete list of scoring options, check https://scikit-learn.org/1.4/modules/model_evaluation.html#scoring

linear_solver : str in {‘lbfgs’, ‘liblinear’, ‘newton-cg’,

‘newton-cholesky’, ‘sag’, ‘saga’}, default=’lbfgs’ Algorithm to use in the optimization problem.

kwargs : dict

Additional keyword arguments to pass to the ModelProbing constructor (e.g. every_n_train_epochs, every_n_val_epochs, prog_bar, …).

fit(X, y)[source]¶

Fit the model according to the given training data and log the classification metrics.

Parameters:

X : {array-like, sparse matrix} of shape (n_samples, n_features)

Training vector, where n_samples is the number of samples and n_features is the number of features.

y : array-like of shape (n_samples,)

Target vector relative to X.

log_metrics(pl_module, y_pred, y_true)[source]¶: Log the metrics given the predictions and the true labels.

predict(X)[source]¶: Predict the probe on new data X.