Spike Timing-Dependent Construction Simulation

Simulation of an STDP-based constructive algorithm for spiking neural networks

Spike timing-dependent construction is an algorithm designed to "grow" a spiking network from an initial population of neurons based on adaptation under Hebbian spike timing-dependent plasticity (STDP). Modelling biological neural network adaptation, STDP is a process of synapse efficacy adaptation from the relative timing of pre- and post-synaptic neuron spikes. Asymmetrical Hebbian STDP produces an increase in excitatory synapse strength when the pre-synaptic neuron spikes immediately prior to the spiking of the post-synaptic neuron.

This constructive algorithm combines STDP with the assumption that simulated neurons exist as a sub-population in a larger neural system with connections to external neurons. In the event that the set of recently active input neurons does not have an associated simulated neuron with equivalent connections, an external neuron with those connections is assumed to exist and spike. Under Hebbian STDP this causes the strengthening of connections to the external neuron and a functional association with the recently active input neurons to form. This external neuron, with connections to the set of active input neurons, is then added to the simulated neurons.

This simulation has been developed as a demonstration of spike timing-dependent construction applied to a randomised 2-dimensional field of neurons. Cyclic motion of a spatial input region causes periodic activation of input neurons and triggers the construction of new neurons. The simulation displays the input neuron locations and the activation and growth of neurons over time.