We quantify the value of classical AM/FM andspectral moments (e.g., amplitude-modulation index, frequencydeviation, spectral kurtosis/skewness) against learned representations in modern RF ensembles. Using a shared dataset interface,we train a tree-based classical stack on hand-engineered featuresand compare to a learned baseline of identical capacity on thesame samples. We provide (i) SHAP analyses over the classicalstack, (ii) … Continue reading AM/FM Handcrafted Features vs. Learned Features in RF Modulation Classification