Linhong Zhu
Application :
1) recommending friends in an social network e.g.
2) suggesting potential hires in a professional network
===============================================================
temporal link prediction : given a sequence of graph snapshots G1,G2......Gt, predict the new links in t+1
the probability of a link depends on the position of their space
==============================================================================
Our model large moves are less likely
contribution :
1) model the temporal link probability of node pairs
2) address algorithmic issuses by developing the standard global BCGD algorithm
3) a local BCGD algorithm , a global BCGD algorithm, an incremental BCGD faster
4) conduct experiments over large real-life graphs, achieve good predictive quality in both sparse and dense networks
the change of the graph
: latent space representation for node set
:a row vector the latent space representation of node u
================================================================================
assumption:
1) temporal smoothness : change gradually
2) network embedding : closer in the original space , closer in the temporal latent space
3) latent homophily : closer → more likely to interact
=====================================================================================
problems:
1) temporal latent space inference
find a latent space representation that minimize the following quadratic loss
(1)
: penalizes node u for suddenly changing the location
: ignore all of the diagonal entries when computing
predict Gt+1 at time t+1
satify
Z1,Z2.....Zt → Zt+q → Gt+1
====================================================================
standard BCGD algorithm
(1) can be decomposed into a linked part and a non-linked part
for each node u at time ,we focus on optimizing :
update Z with the following :