2026-11-01 · America/Los_Angeles · 论文 · #15

Spectral Haar Binarizer: Compact Boolean Contrast Representations for Hyperspectral Classification

Tsetlin Machines offer an interpretable, hardware-friendly alternative to conventional machine learning models, but applying them to continuous hyperspectral measurements requires converting spectral signals into Boolean representations. Standard binarisation methods typically discretise each spectral band independently, potentially discarding important relationships between neighbouring wavelengths. In this paper, we introduce the Spectral Haar Binarizer (SHB), a contrast-based spectral representation method that constructs Boolean predicates from differences between selected spectral…

热度 12.6 / 100；排名与评分保留该期记录。