Model-agnostic methods require access to model internals, while model-specific methods do not.

Model-specific methods are tailored to particular model architectures, while model-agnostic methods can be applied to any model type.

Model-agnostic methods are only applicable to neural networks.

Model-specific methods are always more accurate than model-agnostic methods.

Dimensionality reduction

Counterfactual explanations

SHAP

Data augmentation

A visualization of model structure.

An explanation that shows how changing input features would change the model's prediction.

A method for detecting data drift.

A summary of the most important features in a dataset.

Decision tree rules

Neural network visualization

Model-agnostic methods

Model-specific methods