Interpretability
Interpretability: The degree to which a model’s structure and behavior can be directly understood and analyzed. Linear models and trees can offer more readable rules, while post-hoc techniques analyze complex systems. Interpretability and accuracy are not always in conflict; it depends on the task.
In italiano: InterpretabilityLegge V · Overfitting, bias e collasso dei modelliLinear models and trees can offer more readable rules, while post-hoc techniques analyze complex systems. Interpretability and accuracy are not always in conflict; it depends on the task.
Examples
- Regression coefficients
- Tree rules
- Activation analysis