Configure SVC "shrinking" Parameter

The shrinking parameter in scikit-learn’s SVC class controls whether the shrinking heuristic is used during training.

Support Vector Machines (SVMs) are powerful models for classification and regression. The SVC class is scikit-learn’s implementation of support vector classification.

The shrinking parameter determines whether the shrinking heuristic is used during the SVM optimization process. When enabled, shrinking can significantly speed up training, especially on large datasets.

The default value for shrinking is True. The only other valid option is False, which disables the heuristic.

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score
import time

# Generate synthetic dataset
X, y = make_classification(n_samples=10000, n_features=20, n_informative=10,
                           n_redundant=5, random_state=42)

# Split into train and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train with shrinking enabled and disabled
for shrinking in [True, False]:
    start = time.time()
    svc = SVC(kernel='linear', shrinking=shrinking, random_state=42)
    svc.fit(X_train, y_train)
    y_pred = svc.predict(X_test)
    accuracy = accuracy_score(y_test, y_pred)
    end = time.time()
    print(f"shrinking={shrinking}, Training time: {end - start:.3f}s, Accuracy: {accuracy:.3f}")

Running the example gives an output like:

shrinking=True, Training time: 5.404s, Accuracy: 0.834
shrinking=False, Training time: 30.949s, Accuracy: 0.834

The key steps in this example are:

Generate a large synthetic binary classification dataset
Split the data into train and test sets
Train SVC models with shrinking enabled and disabled
Measure the training time and test accuracy for each model

Some tips and heuristics for setting shrinking:

Shrinking is most beneficial when training on large datasets
If training time is not a bottleneck, leaving shrinking enabled is usually best

Issues to consider:

The impact of shrinking on training speed depends on the size and complexity of the dataset
Disabling shrinking should not significantly affect the model’s accuracy

See Also