Lära Random Search in Practice | Hyperparameter Optimization

Svep för att visa menyn

Random Search is a practical method for hyperparameter tuning that overcomes key limitations of Grid Search.

Grid Search: Tries every possible combination of specified hyperparameters; quickly becomes slow and expensive as the parameter space grows, especially with many or continuous values;
Random Search: Samples a fixed number of random combinations from the parameter space; covers more possibilities with fewer iterations and often finds strong solutions faster.

Use Grid Search for small parameter spaces when you need to test every combination. Choose Random Search for large spaces or when only a few hyperparameters likely affect model performance.


              1234567891011121314151617181920212223242526272829303132333435
            
from sklearn.datasets import load_iris
from sklearn.model_selection import RandomizedSearchCV
from sklearn.neighbors import KNeighborsClassifier
from scipy.stats import randint

# Load dataset
iris = load_iris()
X, y = iris.data, iris.target

# Define the model
knn = KNeighborsClassifier()

# Define hyperparameter distributions
param_dist = {
    "n_neighbors": randint(1, 30),
    "weights": ["uniform", "distance"],
    "p": [1, 2]  # 1: Manhattan, 2: Euclidean
}

# Set up RandomizedSearchCV
random_search = RandomizedSearchCV(
    estimator=knn,
    param_distributions=param_dist,
    n_iter=10,
    cv=5,
    random_state=42
)

# Fit to data
random_search.fit(X, y)

# Show best parameters
print("Best parameters found:", random_search.best_params_)
print("Best cross-validation score:", random_search.best_score_)

Note

Random Search is more efficient than Grid Search when dealing with large parameter spaces, as it avoids evaluating every possible combination and can find strong candidates with fewer iterations.

Var allt tydligt?

Tack för dina kommentarer!

Avsnitt 2. Kapitel 2

Fråga AI

Fråga vad du vill eller prova någon av de föreslagna frågorna för att starta vårt samtal

Avsnitt 2. Kapitel 2