Magnetic domain wall neuron with lateral inhibition

Abstract

The development of an efficient neuromorphic computing system requires the use of nanodevices that intrinsically emulate the biological behavior of neurons and synapses. While numerous artificial synapses have been shown to store weights in a manner analogous to biological synapses, the challenge of developing an artificial neuron is impeded by the necessity to include leaking, integrating, firing, and lateral inhibition features. In particular, previous proposals for artificial neurons have required the use of external circuits to perform lateral inhibition, thereby decreasing the efficiency of the resulting neuromorphic computing system. This work therefore proposes a leaky integrate-andfire neuron that intrinsically provides lateral inhibition, without requiring any additional circuitry. The proposed neuron is based on the previously proposed domain-wall magnetic tunnel junction devices, which have been proposed as artificial synapses and experimentally demonstrated for nonvolatile logic. Single-neuron micromagnetic simulations are provided that demonstrate the ability of this neuron to implement the required leaking, integrating, and firing. These simulations are then extended to pairs of adjacent neurons to demonstrate, for the first time, lateral inhibition between neighboring artificial neurons. Finally, this intrinsic lateral inhibition is applied to a ten-neuron crossbar structure and trained to identify handwritten digits and shown via direct large-scale micromagnetic simulation for 100 digits to correctly identify the proper signal for 94% of the digits.

Maximized lateral inhibition in paired magnetic domain wall racetracks for neuromorphic computing

“…Based on neuron geometry, material and spacing = 90 nm, we estimate the stray field acting on DWI in inhibition case to be = −9 Oe [28]. Compared to the amount of LI shown in [23], here LI is largely maximized by means of optimizing wire interaction strength. These results suggest that for a given neuron geometry, maximum LI is achieved when the magnetic interaction strength coincides with the WB field 4 .…”

Shape-Based Magnetic Domain Wall Drift for an Artificial Spintronic Leaky Integrate-and-Fire Neuron

“…The LIF neuron has also received much interest as a representation of biological neuron behavior, and is a slight modification of the original integrate-and-fire neuron [15]. It emulates the behavior of a biological neuron by receiving and emitting current spikes.…”

“…Previously, we proposed an LIF neuron using a 3T-MTJ device with an additional ferromagnetic layer placed underneath the nanowire track [15]. The ferromagnet produces a constant magnetic field oriented in the -z-direction that causes the DW to gradually travel in the -x-direction.…”

“…As described in [15], current through the device is integrated through motion of the DW. The DW velocity is dependent on the applied current as shown in Fig.…”

“…Leaking: continuous diminishing of the accumulated signal, and Several artificial neurons have previously been proposed based on CMOS [11], magnetoelectric MTJs [12], spin-transfer torque random-access memory [13], and 3T-MTJs [10], [14]. While most prior work requires external circuitry to implement the leaking, we previously proposed an artificial neuron that provided all three behaviors with a single 3T-MTJ device [15].…”