A single magnetic tunnel junction for the implementation of a spiking neuron

We propose a magnetic tunnel junction device to perform bio-realistic neuron firing for spiking neural networks. Spiking neural networks is the third generation of artificial neural networks and promises significantly reduced energy consumption, especially for sparse time-dependent data, by mimicking the biological brain. Magnetic tunnel junctions allow for drop-in replacement in current SNN CMOS-based hardware implementations to increase scalability and efficiency. The magnetization dynamics in the magnetic tunnel junction fully emulates the features of the Huxley-Hodgkin model of the neuron. Hence, the device allows for the implementation of biologically plausible algorithms. The device also emulates the simpler and commonly used Leaky-Integrate and Fire model.