Scikit-Learn RandomizedSearchCV LabelPropagation

Hyperparameter tuning is essential for optimizing machine learning models. In this example, we’ll demonstrate how to use scikit-learn’s RandomizedSearchCV for hyperparameter tuning of a LabelPropagation model, commonly used for semi-supervised learning tasks.

Random search is a method for evaluating different combinations of model hyperparameters. Unlike grid search, it samples a fixed number of hyperparameter combinations from a specified distribution, making it more efficient when searching over a large hyperparameter space.

LabelPropagation is a semi-supervised learning algorithm that spreads labels through the dataset based on the data similarity. It iteratively assigns labels to unlabeled data points by propagating labels from labeled data points through the data structure.

Key hyperparameters for LabelPropagation include the kernel type (kernel), which determines the similarity function used; the parameter for the RBF kernel (gamma), which controls the width of the Gaussian kernel; and the number of neighbors (n_neighbors) for the KNN kernel, which specifies how many neighbors to consider in the propagation.

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split, RandomizedSearchCV
from sklearn.semi_supervised import LabelPropagation
from scipy.stats import uniform, randint

# Generate synthetic dataset
X, y = make_classification(n_samples=1000, n_features=20, n_informative=15, n_redundant=5, random_state=42)

# Set a portion of the labels to -1 to simulate unlabeled data
y[:200] = -1  # Assume the first 200 samples are unlabeled

# 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, stratify=y)

# Define the model
model = LabelPropagation()

# Define hyperparameter distribution
param_dist = {
    'kernel': ['knn', 'rbf'],
    'gamma': uniform(loc=0.1, scale=1.0),  # Only relevant for 'rbf' kernel
    'n_neighbors': randint(3, 7)          # Only relevant for 'knn' kernel
}

# Perform random search
random_search = RandomizedSearchCV(estimator=model,
                                   param_distributions=param_dist,
                                   n_iter=50,
                                   cv=5,
                                   scoring='accuracy',
                                   random_state=42)
random_search.fit(X_train, y_train)

# Report best score and parameters
print(f"Best score: {random_search.best_score_:.3f}")
print(f"Best parameters: {random_search.best_params_}")

# Evaluate on test set
best_model = random_search.best_estimator_
accuracy = best_model.score(X_test, y_test)
print(f"Test set accuracy: {accuracy:.3f}")

Running the example gives an output like:

Best score: 0.721
Best parameters: {'gamma': 0.8319939418114051, 'kernel': 'knn', 'n_neighbors': 3}
Test set accuracy: 0.745

The steps are as follows:

1. Generate a synthetic dataset using scikit-learn’s make_classification function.
2. Simulate unlabeled data by setting a portion of the labels to -1.
3. Split the dataset into train and test sets using train_test_split.
4. Define the LabelPropagation model.
5. Define the hyperparameter distribution with different values for kernel, gamma, and n_neighbors hyperparameters.
6. Perform random search using RandomizedSearchCV, specifying the LabelPropagation model, hyperparameter distribution, 100 iterations, 5-fold cross-validation, and accuracy scoring metric.
7. Report the best cross-validation score and best set of hyperparameters found by random search.
8. Evaluate the best model on the hold-out test set and report the accuracy.

By using RandomizedSearchCV, we can efficiently explore different hyperparameter settings and find the combination that maximizes the model’s performance. This automated approach saves time and effort compared to manual hyperparameter tuning and helps ensure we select the best configuration for our LabelPropagation model.