Spiking Neural Networks: Implementing the Izhikevich Model

In 2003, Eugene Izhikevich proposed an artificial neuron model for Spiking Neural Networks. The model was computationally efficient yet could demonstrate multiple neuron types using a small set of parameters. The implementation of the model in some research is often based on the original implementation. However issues with the order of operations (such as the O(n2) impact of v on u) influence the behaviour of the model. This book includes an improved implementation. Other implementation aspects are included such as: Implementing the model using integers rather than floating point; Using “global variables” to represent “neuromodulators”; Using bitfields to represent spikes; and Use of procedural generation for links between neurons. This book is aimed at people trying to implement the Izhikevich model, however does not contain substantive volumes of source code. It may be of interest to those involved in computational neuroscience and artificial intelligence. It does not contain practical “real-world” applications of the model and should be considered a foundation to build upon.