Competitive learning with spiking nets and spike timing dependent plasticity

This paper explores machine learning using biologically plausible neurons and learning rules. Two systems are developed. The first, for student performance categorisation, uses a two layer system and explores data encoding mechanisms. The second, for digit categorisation, explores competitive behaviour between categorisation neurons using a three layer system with an inhibitory layer. Both are successful. The competitive mechanism from the second system is more plausible biologically, and, by using one neuron per input feature, uses fewer neurons