Graph-Based Global
Reasoning Networks
Yunpeng Chen, Marcus Rohrbach , Zhicheng Yan , Shuicheng Yan, Jiashi
Feng , Yannis Kalantidis
Pre by Guangrui Li
Motivation
• Y = AXW
• W denotes parameters of the convolution kernels d*d
• A is the adjacent Matrix n*n
• X is the node feature n*d
• Torch.enisum(‘xnn,xnd,dd->xnd’)
Graph based global reasoning networks
Graph based global reasoning networks
Graph based global reasoning networks
Graph based global reasoning networks

Graph based global reasoning networks

  • 1.
    Graph-Based Global Reasoning Networks YunpengChen, Marcus Rohrbach , Zhicheng Yan , Shuicheng Yan, Jiashi Feng , Yannis Kalantidis Pre by Guangrui Li
  • 2.
  • 3.
    • Y =AXW • W denotes parameters of the convolution kernels d*d • A is the adjacent Matrix n*n • X is the node feature n*d • Torch.enisum(‘xnn,xnd,dd->xnd’)