Lewis Dyer


Office: Room G111, 18 Lilybank Gardens, Glasgow G12 8RZ

Teams: Lewis Dyer (PGR) [note that messages sent to my staff Teams account are likely to go unread]

Pronouns: he/him


ORCID iDhttps://orcid.org/0000-0002-7685-5860

Research title: Approximate Counting in Graph Databases

Research Summary

My research focuses on counting, sampling and enumeration problems in the context of variable selection problems, which including taking large datasets and selecting a small subset of variables in other to model some value of interest. My research covers a wide variety of areas:

  • Many variable selection methods used in practice are inherently greedy, where variables are only added to a model if they immediately improve said model. Part of my research involves considering the weaknesses of greedy variable selection methods in practice, and identifying where these weaknesses arise in large datasets.
  • Rather than seeking the optimal subset of variables, we may seek to obtain all possible subsets that are "close enough" to optimal - for example, in medical diagnostics, this could allow alternate sets of tests to be performed if a particular test is unavailable (say if it is unsuitable for a particular patient, or if this test currently has a long processing backlog).
  • I also aim to assess the viability of using random walk for variable selection problems, where a weighted graph is constructed between different subsets of the variable set, and a random walk adds and removes variables attempting to find an optimal subset of variables.


I am currently tutoring Algorithms and Data Structures 2, a second year course in algorithmics for computing science students.

Previously, I have been a tutor and/or a demonstrator for courses in discrete mathematics, introductory courses in computer systems and networked and operating systems, and summer schools for incoming first year students in mathematics and computing science.