.png){kind=icon}
.png){kind=icon}
Return
AstN-Net: Rethinking neural connections with astrocyte-driven perspective
.png){kind=icon}
PRE
AI
张
Z
G
S
Y
DOI：10.1016/j.patcog.2026.114537.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Astrocyte-driven neuron model with global integration and bidirectional modulation. • AstN-Net dynamically modulates weights via astrocyte network for structural adaptation. • Alternating training extracts compact sub-networks without extra fine-tuning. • Achieves superior accuracy and efficiency across CNNs, Transformers, and vision tasks.
Keywords:
Neural network
Network structure learning
Astrocyte-driven neuron model
Adaptive modulation of neural connections
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W
Organization
Cited Papers