The project proposal considered two research lines about using evolutionary computation methods for classification tasks bas ed on the nearest neighbour rule.
In the first research line, two separate systems coevolve with each other. The first system is responsible of placing prototypes on the input space. The second system uses Genetic Programming to evolve distance functions, which are then associated to prototypes from the first system.
In the second research line, swarm particle systems are used in a novel way for classification tasks (also with the nearest neighbour rule). The main contribution of this research line is to use the whole population of particles to represent the set of prototypes, instead of having every single particle representing the entire set (the later is the standard approach). The advantages of our approach are the reduction of the search space and the flexibility in the size of the solution