Vasily Shelkov Undergraduate Dissertation 2015/16

Building Interactive Machine Learning Visualizations

Supervised by N.Lawrence

Abstract

Machine learning is becoming a real thing of science fiction in the public's eyes. Recent media reports have caused misunderstanding and, in some cases, fear about what the outcomes of machine learning could become. Also, possibilities of machine learning have frequently been misrepresented.

'Cold reading', in essence, is the statistical analysis of basic information about a given person to produce insights that are revealing to them. A prototype application called the Scikic already exists which acts as a cold reader using a Bayesian Network and Collaborative Filtering.

  The following project will look at visualizing the current Scikic's machine learning techniques. This will be done by implementing a more effective conversational interface than the prototype, where each question answered will visually represent how machine learning is being used to produce personal insights for any of its users.

The final implemented application can be found at http://scikic.org/.