Agreement Disagreement Guided Knowledge Transfer for Cross-Scene Hyperspectral Imaging

TL;DR

The paper tackles cross-scene hyperspectral imaging where transfer learning is hampered by gradient conflicts and a lack of target-feature diversity. It introduces ADGKT, a unified framework that combines agreement (GradVac for gradient direction alignment and LogitNorm for balanced logits) with disagreement (Disagreement Restriction via partial distance correlation and an ensemble with reverse distillation) to robustly transfer knowledge. Extensive experiments on Indian Pines, Pavia, and Houston datasets demonstrate state-of-the-art performance and validate the contributions of each component through ablations. The work offers a practical, scalable approach to improve generalization in heterogeneous HSI transfers by addressing both optimization dynamics and target-information preservation.

Abstract

Knowledge transfer plays a crucial role in cross-scene hyperspectral imaging (HSI). However, existing studies often overlook the challenges of gradient conflicts and dominant gradients that arise during the optimization of shared parameters. Moreover, many current approaches fail to simultaneously capture both agreement and disagreement information, relying only on a limited shared subset of target features and consequently missing the rich, diverse patterns present in the target scene. To address these issues, we propose an Agreement Disagreement Guided Knowledge Transfer (ADGKT) framework that integrates both mechanisms to enhance cross-scene transfer. The agreement component includes GradVac, which aligns gradient directions to mitigate conflicts between source and target domains, and LogitNorm, which regulates logit magnitudes to prevent domination by a single gradient source. The disagreement component consists of a Disagreement Restriction (DiR) and an ensemble strategy, which capture diverse predictive target features and mitigate the loss of critical target information. Extensive experiments demonstrate the effectiveness and superiority of the proposed method in achieving robust and balanced knowledge transfer across heterogeneous HSI scenes.

Figures (3)

• Figure 1: Our method contains two mechanisms including agreement and disagreement during the knowledge transfer from source scene to target scene. The agreement mechanism contain GradVac to alleviate gradient conflict and LogitNorm to mitigate the dominant gradient. Additionally, the disagreement mechanism comprise disagreement restriction to promote diversity and ensemble to capture agreement and disagreement aspects of target features.
• Figure 2: The agreement mechanism is proposed in our model. To alleviate gradient conflict and dominant gradient issues in shared encoder $G$, we employ GradVac to adapts the gradients $g_s$ to $g_s'$, thereby reducing gradient conflict. In addition, we utilize LogitNorm to get updated logits $\hat{z_s}$ and $\hat{z_t}$ , controlling their magnitude to prevent potential underfitting.
• Figure 3: The disagreement mechanism is introduced in our model. To prevent the risk of losing critical target features, we employ disagreement restriction through $E_{\text{DiR}}$ to promote diversity. Then, we utilize the ensemble component through $E_{\text{en}}$ to capture a diverse set of target features.