Hardware realization of the multiply and accumulate operation on\n radio-frequency signals with magnetic tunnel junctions

Abstract

Artificial neural networks are a valuable tool for radio-frequency (RF)\nsignal classification in many applications, but digitization of analog signals\nand the use of general purpose hardware non-optimized for training make the\nprocess slow and energetically costly. Recent theoretical work has proposed to\nuse nano-devices called magnetic tunnel junctions, which exhibit intrinsic RF\ndynamics, to implement in hardware the Multiply and Accumulate (MAC) operation,\na key building block of neural networks, directly using analogue RF signals. In\nthis article, we experimentally demonstrate that a magnetic tunnel junction can\nperform multiplication of RF powers, with tunable positive and negative\nsynaptic weights. Using two magnetic tunnel junctions connected in series we\ndemonstrate the MAC operation and use it for classification of RF signals.\nThese results open the path to embedded systems capable of analyzing RF signals\nwith neural networks directly after the antenna, at low power cost and high\nspeed.\n