Learning with Subset Stacking (LESS)

LESS is a supervised learning algorithm that is based on training many local estimators on subsets of a given dataset, and then passing their predictions to a global estimator. You can find the details about LESS in our manuscript.

Installation

Install LESS from PyPI:

pip install less-learn

Getting Started

Below is a simplified example of how to use LESS

from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error
from less import LESSBRegressor

# Generate a synthetic regression dataset
X, y = make_regression(n_samples=1000, n_features=20, random_state=42)

# Split into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.3, random_state=42
)

# Initialize and train the LESS model
less_model = LESSBRegressor(random_state=42)
less_model.fit(X_train, y_train)

# Make predictions and evaluate performance
y_pred = less_model.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
print(f"Test MSE of LESS: {mse:.2f}")

Note: LESS employs Euclidean distances combined with a radial basis function (RBF) kernel by default. It is therefore highly recommended to scale or normalize input features prior to model training to ensure optimal performance.

Citation

Our software can be cited as:

  @misc{LESS,
    author = "Ilker Birbil and Samet Copur",
    title = "LESS: Learning with Subset Stacking",
    year = 2025,
    url = "https://github.com/sibirbil/LESS/"
  }