kan-ftir-food-adulteration: interpretable Kolmogorov–Arnold networks for FTIR food-adulteration analysis

Leakage-free, equal-budget pipeline benchmarking an interpretable Kolmogorov-Arnold network against PLS, SVR, random forest, MLP and a 1-D CNN for FTIR-based detection and quantification of adulteration in olive oil, coffee and fruit juice, with closed-form symbolic equations and chemically meaningful response curves. Ships with the three open ATR-FTIR datasets (CC-BY-4.0).