Home / Publications
One arc —— Structural Reliability → Scalability → Selective Observation → Interpretability
My work circles a single question: how do graph learning systems stay reliable at real scale, under limited observation, and with complex semantics?
Structural Reliability
Node, edge and community constraints in one program, suppressing the structural degeneracy that overlapping community detection falls into.
Zihe Zhou, Samin Aref · Triad: Suppressing Structural Degeneracy in Overlapping Community Detection
Scaling Reliable Structure
Scaling sparse structural backbones to large networks, for overlapping community detection that stays both reliable and fast.
Zihe Zhou, Samin Aref · Overlapping Network Community Detection Using Sparse Backbones
Interpretable Generation on a Graph
DSEG-Char: a character-level poetry model where every character is a node in a 5,000-node graph. Writing a line is a walk over it, and each step reports which term — state, edge or context — selected that character.
“Graph learning systems that are more reliable let complex relationships be seen, understood, and trusted.”
Zihe Zhou