search

ICLab4DL / GGN

GGN model for seizure classification (datasets: TUH EEG seizure TUSZ 1.5.2)
MIT License
34 stars 6 forks source link

About update_A(self, mu, sig, pi) function #3

Closed snowsj closed 7 months ago

snowsj commented 8 months ago

*S = mu_k + nsig_k (line 324 in th gnn.py), Is this operation to obtain the updated probability distribution of node connections? Is my understanding incorrect? Why do we need to operate like this? And Sim=torch. einsum ('bnc, bcm ->bnm ', S, S. transfer (2,1))**, what is this calculated for? Thank you for your answer！

zhengdaoli commented 8 months ago

i-th row of S is a random vector of node i, P is the "probability distribution of node connections"