profile picture

Paulius Dilkas

I am currently a Postdoctoral Fellow at the University of Toronto, working under the supervision of Kuldeep S. Meel. Before this, I held a position as a Research Fellow at the National University of Singapore, also under the guidance of Kuldeep S. Meel. I earned my PhD in Robotics and Autonomous Systems from a joint program between the University of Edinburgh and Heriot-Watt University. My doctoral supervisor was Vaishak Belle. Additionally, I obtained an MSci degree in Computing Science from the University of Glasgow.

My research focuses on discrete algorithms that count and compute sums of products based on a logical problem description. Specifically, I study weighted model counting—a weighted version of the #SAT problem—and first-order model counting. To solve these problems, I employ algorithmic techniques such as dynamic programming, knowledge compilation, and different representations of Boolean and pseudo-Boolean functions. Efficiently solving model counting problems is of utmost importance in various domains of artificial intelligence (AI), including explainable AI, neuro-symbolic AI, probabilistic programming, and statistical-relational AI. Moreover, these techniques find applications in bioinformatics, data mining, natural language processing, prognostics, and robotics.

In addition to my primary research interests, I am interested in graph algorithms, constraint satisfaction, and search algorithms. Previously, I have worked on projects involving graph algorithms, algorithm portfolios, formal modelling using bigraphs, and inverse reinforcement learning.

Publications

FOMC
Paulius Dilkas, Vaishak Belle. Synthesising Recursive Functions for First-Order Model Counting: Challenges, Progress, and Conjectures. KR 2023
paper video slides code
WMC
Paulius Dilkas. Generating Random Instances of Weighted Model Counting: An Empirical Analysis with Varying Primal Treewidth. CPAIOR 2023
paper slides code
PBP
Paulius Dilkas, Vaishak Belle. Weighted Model Counting Without Parameter Variables. SAT 2021
paper video slides code
CW
Paulius Dilkas, Vaishak Belle. Weighted Model Counting with Conditional Weights for Bayesian Networks. UAI 2021
paper supplement video slides poster code
random
Paulius Dilkas, Vaishak Belle. Generating Random Logic Programs Using Constraint Programming. CP 2020
paper video slides code

Preprints, Theses, etc.

handbook
Mapping the Neuro-Symbolic AI Landscape by Architectures: A Handbook on Augmenting Deep Learning Through Symbolic Reasoning (with Jonathan Feldstein, Vaishak Belle, and Efthymia Tsamoura, 2024)
preprint
thesis
Generalising Weighted Model Counting (supervised by Vaishak Belle, 2023)
PhD thesis slides
VIGPIRL
Variational Inference for Inverse Reinforcement Learning with Gaussian Processes (supervised by Bjørn Sand Jensen, 2019)
MSci report slides code
NBRS
Nondeterministic Bigraphical Reactive Systems for Markov Decision Processes (supervised by Michele Sevegnani, 2018)
internship report slides code
MCS
Algorithm Selection for Maximum Common Subgraph (supervised by Ciaran McCreesh and Patrick Prosser, 2018)
BSc dissertation slides code
GED
Clique-Based Encodings for Graph Edit Distance (supervised by Ciaran McCreesh, 2017)
internship report slides code

Software

...
Crane

A model counter for first-order logic

code

Teaching

2019-2022
At the School of Informatics, University of Edinburgh, I worked on the following courses:
2017-2019
At the School of Computing Science, University of Glasgow I worked as a demonstrator for the following courses:
2012-2017
Before that, I worked as a Distance Learning Teacher in Mathematics at the National Student Academy in Lithuania

Awards

  • 2023
    • UAI 2023 Top Reviewer
  • 2019
    • 3-Year Scholarship from the EPSRC CDT in Robotics and Autonomous Systems
    • MSci Class Prize
  • 2018
    • EPSRC Vacation Scholarship
    • Level 4 Project with Best Product
  • 2017
    • EPSRC Vacation Scholarship
    • Level 3 Honours Class Prize for Computing Science
  • 2016
  • 2015
    • O'Reilly Academic Prize for Best Overall Performance in Assessed Coursework in Level 1 Computing Science
    • Lorimer Bursary Prize