Huber Regression
Robust linear regression resistant to outliers
Huber Regression minimizes the Huber loss, which behaves like squared error for small residuals and absolute error for large residuals. This makes it resistant to outliers while remaining efficient for well-behaved data.
When to use:
- Regression datasets with outliers that would distort standard linear regression
- When a robust linear model is preferred over tree-based approaches
- Financial, sensor, or measurement data with occasional extreme values
Input: Tabular data with the feature columns defined during training Output: Continuous predicted value
Model Settings (set during training, used at inference)
Epsilon (default: 1.35) Threshold separating quadratic (inlier) from linear (outlier) loss. Lower values treat more points as outliers.
Max Iter (default: 100) Maximum iterations for the solver.
Alpha (default: 0.0001) L2 regularization strength.
Fit Intercept (default: true) Whether to include a bias term.
Inference Settings
No dedicated inference-time settings. Predictions use the trained robust coefficients.