Zihe Zhou
Zihe Zhou

Zihe Zhou (周子贺)

Researcher in graph-structured learning & overlapping community detection · PhD applicant, Fall 2027

Email GitHub ORCID

I build fast, scalable algorithms and systems for graphs. My flagship work, Highway, detects overlapping communities in networks up to 1.13 million nodes using sparse backbones — and is now integrated into cdlib, a mainstream community-detection library.

I'm currently finishing my MEng at the University of Toronto and applying to CS/EECS PhD programs for Fall 2027. I'm drawn to graph research that is both scalable and interpretable, and to the connection between graph theory and real-world problems.

News

Publications

Highway

Overlapping Network Community Detection Using Sparse Backbones

Zihe Zhou, Samin Aref

ASONAM 2026 (Springer proceedings) · Accepted

Case study → arXiv Benchmark cdlib

A four-step sparse-backbone method. Evaluated on ~3,000 synthetic graphs and three real SNAP networks (up to 1.13M nodes / 2.99M edges); the only method to finish all three within 300s (7.34× faster on the largest instance).

Triad

Triad: Suppressing Structural Degeneracy in Overlapping Community Detection

Zihe Zhou, Samin Aref

WAW 2026 · Presentation

Case study → Talk video Slides

A QCP formulation with node/edge/community constraints that explicitly suppresses structural degeneracy — the predecessor method that led to Highway.

Selected Research

One arc — structural reliability → scalability → interpretability.

Origin

Triad

Structurally reliable overlapping assignments through node/edge/community constraints — precise but hard to scale.

Read the case study → Flagship

Highway

Scalable overlapping community detection through sparse structural backbones — to 1.13M nodes.

Read the case study →
Future

DSEG

Carrying overlapping-community ideas toward interpretable, graph-structured semantic control — exploratory.

Selected System

HELM

A full-stack AI wealth operating system

A six-layer data platform with a governed data spine and multiple clients, grown out of a finance-ML course project. Evidence of engineering maturity rather than a research thrust.

Talks & Presentations

Background

University of Toronto — MEng, Mechanical & Industrial Engineering (Data Analytics & ML), 2025–2026. Research advised by Prof. Samin Aref. Four graduate courses completed — all A+.

University of Waterloo — B.C.S. Honours with Distinction; CS major, AI specialization, Computational Math minor, 2020–2025.