Jason Li

Research Interests

My research applies modern algorithmic techniques to the hardest, longstanding open problems in the field of fast graph algorithms. To illustrate, here are some of my recent discoveries:

• the first almost-linear time algorithm for deterministic global minimum cut through the expander decomposition technique,
• the first almost-linear time all-pairs minimum cut (Gomory-Hu tree) algorithm through the isolating cuts framework and recent advances in dynamic graph algorithms, and
• the first nearly work- and time-optimal parallel algorithm for approximate shortest path through methods in continuous optimization.

These modern techniques are all variations of two recurring themes, preconditioning and locality, which I explore in my thesis below. Loosely speaking, I see preconditioning and locality as reductions from worst-case instances to well-behaved and local instances, respectively.

Thesis: Preconditioning and Locality in Algorithm Design
EATCS Distinguished Dissertation Award (2021)
[slides]

Selected Publications (show all)

All Publications (show selected)

Preprints

Contact

Office: Room 120, Simons Institute, UC Berkeley

Email: jm[my last name]@alumni.cmu.edu