Project

The course has a project. It requires you to build a computational model of categorization.

Project description

The problem of categorization (or classification or concept formation) is fundamental to cognition. Categorization organizes sensory input into categories or groups which are 'similar' and often we give it a name - for example 'human', 'dog' or 'neem tree'. And one central question in cognition, that is yet to be answered, is how do humans (or animals) categorize. There are many (approx. a dozen) theories on how humans categorize.

The project requires you to pick one of the theories/formal models of categorization (see below) and build a computational model for it. Categorization can be supervised or unsurpervised and some models may handle only one kind of categorization. There is even a debate on whether the two kinds are distinct or just variants of a single latent process.

The models from which you can choose are available in the book:
E M Pothos, A J Wills (Eds), Formal Approaches in Categorization, Cambridge Univ. Press, 2011.

Each article in the book that discusses a particular model also contains more references that are relevant.

In your project you have to write a program that implements the model of your choice and then show how the model explains some actual human data and/or predicts some behavioural data that is present in the literature and/or suggests a specific behavioural outcome that the model predicts that is new and not in the literature. You are also expected to be thoroughly familiar with the model you are implementing.

Project time-line

Project demo and presentation on 10th Nov.

Project presentation schedule

The schedule is here

Project evaluation

Project presentation - 10%

Total marks of project component - 10%