Zeru Zhu
I'm Zeru Zhu, a PhD student in Applied Mathematics at Stony Brook University. I enjoy seeing connections between different areas of mathematics.
Recently, I've been interested in AI for mathematics, spectral graph theory, representation theory, and the Langlands program. This site is a growing gallery of my work.
Map of work
Different problems, connected ideas.
One graph of shared questions, methods, and applications. Connected work stays close, with reinforcement learning linking several research directions. Each node represents a project; a project can contain more than one paper.
Hover or focus a node to trace its connections. Select a node to open the project. On small screens, scroll sideways to explore the whole graph.
One graph, with connected work close together and less-connected branches toward the outside. Node sizes are equal.
Read all connections
Algebraic connectivity ↔ MerLean-Prover
MerLean generated the Lean formalization of the extremal connectivity result.Decentralized bandits ↔ Push-sum analysis
Both analyze learning or optimization over directed communication networks; this is a shared setting, not a claim that one implements the other.Push-sum analysis ↔ Distributed VQLS
Both coordinate local optimization variables through neighbor communication.Distributed VQLS ↔ Quantum gate control
Variational quantum circuits and quantum gate control address different parts of quantum computation.Network SEIRV ↔ SIS minimum effort
Both model spreading processes and interventions in networked populations.Algebraic connectivity ↔ Quantum gate control
Both use reinforcement learning to choose actions: graph-edge additions and quantum-control actions.Algebraic connectivity ↔ Cooperative locomotion
Both study learned policies and rewards, applied to graph construction and cooperative motion.Quantum gate control ↔ Cooperative locomotion
Both use reinforcement learning for control, in quantum dynamics and multi-agent locomotion.Algebraic connectivity ↔ TSP policy learning
Both learn sequential choices that construct a discrete graph solution under constraints.Quantum gate control ↔ TSP policy learning
Both learn sequential decision policies, for quantum control and tour construction.Cooperative locomotion ↔ TSP policy learning
Both investigate how reward and policy learning shape sequences of decisions.
Inside the gallery
The details behind the work.
Project pages bring together the question, mathematical ideas, methods, and evidence behind each piece of work, with room for more detail than a paper list or CV.
- 01
Question
The problem and why it is worth investigating.
- 02
Ideas
The mathematical formulation, insight, and connections.
- 03
Method
The approach taken and my contribution to the work.
- 04
Evidence
Proofs, experiments, limitations, and questions still open.
Current interests
AI for mathematics, spectral graph theory, representation theory, and the Langlands program—and the insights that can emerge between mathematical viewpoints.