Optimisation of the Topology of a Multi-Layer Perceptron

Artificial intelligence is becoming increasingly present in our society. To this end several strategies have been created. Artificial Neural Networks was one of them and has several architectures and topologies. The Multiple Perceptron for example is a neural network that has a great capacity for generalisation that is when used for pattern classification it is able to correctly classify samples that have never been presented to it using only its experience with previous classifications. However the generalisation capacity of the perceptron is proportional to the quality of its topology i.e. good generalisation requires good topology. However finding the ideal topology for a perceptron is not a simple task. This work analyses the metrics used to find the best topology for a given problem.