Regression
Mean absolute error
Implemented
regression.maeDefinition
Average absolute difference between predictions and targets, in target units.
Formula
MAE = (1/N) Σᵢ |yᵢ − ŷᵢ|
Range: [0, ∞)
Inputs and outputs
- y_true: array of real targets
- y_pred: array of real predictions
- sample_weight: optional
Returns: float
Assumptions
No assumptions beyond valid, aligned inputs of the documented types.
Limitations
No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.
Python API
import evalsuite as es
es.mae(y_true_r, y_pred_r)References
- Willmott, C. J., & Matsuura, K. (2005). Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Research, 30(1), 79–82.