Overview

The goal of the project is to evolve user interface layouts. Given a reference interface layout, we would like the genetic algorithm to evolve an initial population of random layouts towards the reference layout.

Eventually we would like to evolve complex user interfaces and to learn the characteristics of the resulting interfaces. By doing so we hope to eventually define a set of characteristics which can be associated with aesthetics. We want to learn good and bad user interface layouts.

Source Code

The source code can be obtained from the following link: UI Evolution
The code is given as is, so use at your own discretion.

Results

Proof of concept

User Feedback

The GA produces a random initial population of interface layouts. The user provides feedback by selecting the best and worst individuals in the population. The rest of the individuals get their fitness by creating an interpolation based on the best and worst individuals.
Below we see the initial population of random user interface layouts and the converged population of interface layouts after several generations.


   

This shows the interpolation devised from the fitness of the best and worst individuals.