Infinitely Deep Graph Transformation Networks

Dec 12, 2023ยท
Lei Zhang
Lei Zhang
,
Qisheng Zhang
,
Zhiqian Chen
,
Yanshen Sun
,
Chang-Tien Lu
,
Liang Zhao
ยท 1 min read
Type
Publication
In 2023 IEEE International Conference on Data Mining
minโกฮ˜v,ฮ˜e,Wev,Wve,ฮฉL1(Fโ€ฒ,gฮ˜e(Hv))+L2(Eโ€ฒ,fฮ˜e(He)),subject to: He=ฯ•(WveHvB),Hv=ฯ•(Wevฯ•(WveHvB)BโŠบ+bฮฉ(E0,F0)),ฮปpf(BBโŠบ)ฮปpf(Wevโˆฃโ‹…โˆฃWveโˆฃโ‹…โˆฃ)<1. \min_{\Theta_v, \Theta_e, W_{ev}, W_{ve}, \Omega} \mathcal{L}_1(F', g_{\Theta_e}(H_v)) + \mathcal{L}_2(E', f_{\Theta_e}(H_e)) , \\ \text{subject to: } H_e = \phi(W_{ve} H_v B ) , \\ H_v = \phi( W_{ev} \phi( W_{ve} H_v B ) B^{\intercal} +b_{\Omega}(E_0, F_0)) ,\\ \lambda_{pf}(B B^{\intercal}) \lambda_{pf} ( W_{ev}^{|\cdot|} W_{ve}^{|\cdot|}) <1 .
Graph Neural Network
Lei Zhang
Authors
Lei Zhang
Assistant Prof.