Neuromorphic Networks of Spiking Neurons
This chapter describes some general properties of neuromorphic systems, and how neurons and synapses can be implemented in the complimentary metal-oxide semiconductor electronic medium using hybrid analog/digital very large scale integrated (VLSI) technology. The primary feature of the vast majority of conventional VLSI circuits is that they are purely digital: They use transistors as on-off switches, and they represent numbers as collections of binary digits. Synapses are highly specialized structures that, by means of complex chemical reactions, allow neurons to transmit signals to target neurons. Most neuromorphic circuits are designed to emulate populations of spiking neurons and so comprise massively parallel arrays of silicon synapses and neurons. Biological neurons communicate with one another through dedicated point-to-point axons that must ramify widely to make the necessary connections between a source neuron and its many target synapses. Integrate and fire neurons integrate presynaptic input currents and generate a voltage pulse analogous to an action potential when the integrated voltage reaches a threshold.