Striking Features for Similarity Search

  • Striking Features for Similarity Search

Striking Features for Similarity Search

Principal Supervisor: Deepak Padmanabhan

+ Project Description

Many of us who work in the QUB Computer Science Building (CSB) find it easy to describe our workplace to any Belfast resident – you would most likely just need to say “the multi-colour glass walled building on Malone Road”. This is so since the building appears to be so very striking within its context. First, there are very few Belfast buildings that sport multiple colours, and second, there are very few university buildings that are have a lot of glass-walls. Humans find it naturally easy to unambiguously describe entities by way of features that appear striking. However, from the traditional attribute-value representation of data, it is easy to regard QUB CSB as similar to buildings in its vicinity. There are way too many 3-storey buildings, of which QUB CSB is one, and there are way too many large buildings in the university area, CSB is large too. Thus, it is not the number of matching attributes between CSB and other buildings that give its uniqueness, but the striking nature of the values it takes on two attributes, colour and wall-type. Similarity Search systems that compare entities such as buildings based on the similarity among their attributes, and aggregate such similarities to arrive at a quantification of similarity would naturally not be able to account for “striking” features. This project explores a paradigm change in similarity search, one where striking features are mined and explicitly modelled, in an effort to get similarity search systems to imbibe the human mind’s notion of similarity that makes plentiful usage of striking features. 

+ How to Apply

Applicants should apply electronically through the Queen’s online application portal at:

+ Contact Details

Supervisor Name: Dr. Deepak Padmanabhan

03-032 CSB      



+44 (0)28 9097 4874