Regression
Root mean squared log error
Implemented
regression.rmsleDefinition
RMSE computed on log(1 + value), emphasising relative rather than absolute error.
Formula
RMSLE = √((1/N) Σᵢ (log(1 + yᵢ) − log(1 + ŷᵢ))²)
Range: [0, ∞)
Inputs and outputs
- y_true: array of real targets
- y_pred: array of real predictions
- sample_weight: optional
Returns: float
Assumptions
- Targets and predictions must be greater than −1; invalid domains raise a validation error.
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.rmsle(y_true_r, y_pred_r)References
Primary references for this entry have not been added yet.