Słownik AI
Kompletny słownik sztucznej inteligencji
Local Interpretation Methods
Techniques explaining individual model decisions for specific predictions such as LIME and SHAP.
Global Interpretability
Approaches that allow understanding the model's overall behavior across the entire dataset.
Feature Attribution
Methods quantifying the importance of each input variable in the model's final decision.
White Box Models
Intrinsically interpretable algorithms such as decision trees, linear regressions, and logical rules.
Decision Visualization
Graphical techniques representing decision-making processes and relationships between variables.
Counterfactual Explanations
Hypothetical scenarios showing how to modify inputs to change the model's prediction.
Interpretability of Neural Networks
Specialized methods for understanding and visualizing the decisions of complex deep learning models.
Interpretability Evaluation Metrics
Metrics quantifying the quality, fidelity, and usefulness of explanations generated by models.
Causal Interpretability
Approaches identifying cause-effect relationships rather than mere correlations in AI decisions.
Post-hoc vs. Intrinsic Explanations
Distinction between explanations added after training and those integrated directly into the model's architecture.
Interpretable Rule Extraction
Techniques converting complex models into sets of logical rules understandable by humans.
Interpretability for Regulatory Audit
Methods adapted to legal compliance requirements such as GDPR and transparent AI directives.