Electromagnetic and Laplace domain analysis of memristance and associative learning using memristive synapses modeled in SPICE

The unique properties of memristors could possibly be used in non-volatile memories and neuromorphic computing to drastically reduce area and power dissipation. In this paper, an attempt is made to understand the concept of memristance from an electromagnetic theory perspective and derive an expression for memristance in Laplace domain involving only fundamental material properties. Further, a parameterized SPICE model for the memristor is shown to mimic a synapse in a typical neural network. An ultra-low power and compact neural network is constructed using memristors and the leaky-integrate-fire neuron model to demonstrate associative learning. This shows promise that memristive neuromorphic computing has potential to achieve the ultimate challenge of mimicking the human brain.