.png){kind=icon}
.png){kind=icon}
Return
InsNet: Deep indefinite spectral kernel network
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2026.113337.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Propose InsNet, a deep indefinite spectral kernel network, constructed by stacking indefinite spectral kernel mappings with a novel proposed initialization scheme. • Enable hierarchical modeling of reciprocal and multi-scale data relationships. • Provide a rigorous theoretical analysis of InsNet, examining its architectural components and generalization properties. • Extensive experiments on synthetic and real-world data demonstrate that our InsNet is totally superior to SOTA deep spectral kernel models.
Keywords:
InsNet
indefinite spectral kernel
hierarchical modeling
multi-scale data
deep learning