Associative memories using complex-valued Hopfield networks based on spin-torque oscillator arrays

Abstract Simulations of complex-valued Hopfield networks based on spin-torque oscillators can recover phase-encoded images. Sequences memristor-augmented inverters provide tunable delay elements that implement complex weights by phase shifting the oscillatory output oscillators. Pseudo-inverse training suffices to store at least 12 images in a set 192 oscillators, representing 16 × pixel The energy required an image depends desired error level. For and circuitry considered here, 5% root mean square deviations from ideal require approximately 5 μ s consume roughly 130 nJ. show network functions well when resonant frequency be tuned have fractional spread less than 10 −3 , depending strength feedback.