SMX#
Spectral Model eXplainer
Explain spectral ML models with zone-level predicates
SMX is a global, model-agnostic explainer for spectral classifiers. It turns spectra into expert zones, generates logical predicates, and ranks them with a graph-based centrality score.
What SMX provides#
End-to-end pipeline via the
SMXclassZone extraction and PCA aggregation tailored to spectral inputs
Quantile predicates with bagging and perturbation-based ranking
Directed predicate graph with Local Reaching Centrality (LRC)
Natural-scale threshold reconstruction and Plotly visuals
Faithfulness evaluation with progressive masking
Run SMX end to end on a synthetic dataset in minutes.
Understand how zones, predicates, and graphs fit together.
Explore the full suite of SMX visualization helpers.
Auto-generated reference for every public module.
Scripts and notebooks shipped with the repository.
A guided tour of the SMX source tree.
Contents#
User Guide
API Reference