More about HKUST
Parameterized Algorithms and Complexity Results for Wiener Index and Liquid Democracy
The Hong Kong University of Science and Technology
Department of Computer Science and Engineering
PhD Thesis Defence
Title: "Parameterized Algorithms and Complexity Results for Wiener Index and
Liquid Democracy"
By
Mr. Pavel HUDEC
Abstract:
Decades of algorithmic research have shown numerous natural problems in
computer science to be intractable for general input instances. Naturally, the
question was posed whether the worst-case behavior and hardness results were
realistic, or whether the real-world instances possess exploitable structure
for efficiency. In this thesis, we study graph sparsity parameters that
capture such structure, e.g., the treewidth and vertex cover number of a
graph.
This thesis follows this line of work in two settings: First, it studies
computational chemistry, specifically the computation of the Wiener index of a
chemical molecule. Wiener index is a classical topological index, initially
defined to analyze the boiling point of alkanes and later found applications
in drug synthesis. We experimentally show that molecules have a bounded
treewidth and use this property in designing a faster parameterized
approximation algorithm for computing the Wiener index of a molecule. We
further show that our algorithm achieves a significant speed-up in practice.
We also show treewidth is broadly applicable in computational chemistry, in
particular in computation of classical topological indices, such as the number
of Kekulé structures, the Merrifield-Simmons index, and the Hosoya index.
We provide practical fixed-parameter tractable algorithms for these problems,
and show that they outperform the state-of-the-art often by orders of
magnitude. Finally, we design the fastest known algorithm for the Minimum
Cycle Basis (MCB) problem in graphs of bounded treewidth.
For the second part, we study the problem of computing the success probability
in Liquid Democracy, a classical and widely-studied form of voting with
delegation. We prove that the problem is #P-complete in its most general case.
Moreover, we provide a practical near-cubic algorithm for trees, and extend it
to an XP algorithm for graphs of bounded treewidth. Independently, we design
an FPT algorithm for graphs of bounded vertex cover number.
Date: Tuesday, 22 September 2026
Time: 4:00pm - 6:00pm
Venue: Room 3494
Lifts 25/26
Chairman: TBC
Committee Members: Dr. Jiasi SHEN (Supervisor)
Dr. Amir GOHARSHADY (Co-supervisor, University of Oxford)
Dr. Sunil ARYA
Prof. Dimitrios PAPADIAS
Dr. Maximilian NITZSCHNER (MATH)
Prof. Sandi KLAVŽAR (University of Ljubljana)