White-Box mHC: Electromagnetic Spectrum-Aware and Interpretable Stream Interactions for Hyperspectral Image Classification

2026-01-22

AI Summary

This work addresses the limited interpretability of deep models in hyperspectral image classification due to their black-box nature. To this end, we propose the ES-mHC framework, which uniquely integrates electromagnetic spectrum awareness with a white-box hyper-connection mechanism. By employing structured directional matrices, our approach explicitly models asymmetric information flow among grouped spectral bands and decouples feature representation from interaction structure. This design enables visualization and analysis of internal information pathways, facilitating a shift from black-box to partially white-box modeling. Experimental results demonstrate that the learned hyper-connection matrices exhibit consistent spatial patterns and accelerate the formation of structured interactions as the expansion ratio increases, thereby enhancing model interpretability without compromising classification performance.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

In hyperspectral image classification (HSIC), most deep learning models rely on opaque spectral-spatial feature mixing, limiting their interpretability and hindering understanding of internal decision mechanisms. We present physical spectrum-aware white-box mHC, named ES-mHC, a hyper-connection framework that explicitly models interactions among different electromagnetic spectrum groupings (residual stream in mHC) interactions using structured, directional matrices. By separating feature representation from interaction structure, ES-mHC promotes electromagnetic spectrum grouping specialization, reduces redundancy, and exposes internal information flow that can be directly visualized and spatially analyzed. Using hyperspectral image classification as a representative testbed, we demonstrate that the learned hyper-connection matrices exhibit coherent spatial patterns and asymmetric interaction behaviors, providing mechanistic insight into the model internal dynamics. Furthermore, we find that increasing the expansion rate accelerates the emergence of structured interaction patterns. These results suggest that ES-mHC transforms HSIC from a purely black-box prediction task into a structurally transparent, partially white-box learning process.

Problem

• hyperspectral image classification
• interpretability
• black-box models
• electromagnetic spectrum
• feature interaction

Innovation

• white-box
• electromagnetic spectrum-aware
• hyper-connection
• interpretable interactions
• hyperspectral image classification