Spin Torque Nano-Oscillators as candidates for Artificial Neural Networks

Abstract

Artificial Neural Networks have been widely used with great success for tasks such as input classification. However, they require considerable computing resources. Spin Torque Nano-Oscillators are nanometric devices capable of converting an spin-polarized current into a magnetic oscillation, through the spin-transfer-torque effect. We show that this devices are capable of non-linear behavior such as synchronization, and that their oscillation can be finely adjusted, making them good candidates for effcient, hardware-built neural networks