An Associative Memory with oscillatory CNN arrays using spin torque oscillator cells and spin-wave interactions architecture and End-to-end Simulator

Abstract

An Associative Memory is built by three consecutive components: (1) a CMOS preprocessing unit generating input feature vectors from picture inputs, (2) an AM cluster generating signature outputs composed of spintronic oscillator (STO) cells and local spin-wave interactions, as an oscillatory CNN (O-CNN) array unit, applied several times arranged in space, and (3) a classification unit (CMOS). The end to end design of the preprocessing unit, the interacting O-CNN arrays, and the classification unit is embedded in a learning and optimization procedure where the geometric distances between the STOs in the O-CNN arrays play a crucial role. The O-CNN array has an input vector as a 1D array of oscillator frequencies, and the synchronized O-CNN array codes the output as the phases of the output 1D array. The typical O-CNN array has 1-3 rows of STOs. Simplified STO and interaction macro models are used. A typical example is shown using an End-to-end Simulator.

Spin-torque building blocks

“…In practice, associative memory processors are important building blocks for applications such as pattern recognition and, in general, classification. These considerations, combined with the tiny size and frequency tunability of spin-torque oscillators, have motivated the proposal of associative memory hardware based on arrays of interacting harmonic spin-torque nano-oscillators [110] [111]. It has been shown that these systems can be decomposed in clusters of reasonable size for the arrays of synchronized spin-torque oscillators without performance degradation [112].…”

Optimizing magnetodipolar interactions for synchronizing vortex based spin-torque nano-oscillators

“…A possible solution to these issues could be the synchronization of a few STNOs [6,19,21,22,26,28]. Synchronization among multiple auto-oscillators can also be useful in the framework of developing associative memories architectures [29][30][31][32]. Previous studies reported on synchronization of STNOs interacting with others via spin waves [25,26,33], exchange coupling [6], electric currents [3,28,34], noisy current injection [19], or via magnetodipolar interaction [21,22,[35][36][37][38][39][40][41].…”

Boolean Computation Using Self-Sustaining Nonlinear Oscillators

“…SSNOs exhibit interesting dynamical properties -for example, synchronization [19][20][21] and pattern formation [22][23][24][25] can result when they are coupled together. Biological phenomena such as the synchronized flashing of fireflies [26], circadian rhythms [27,28] and epilepsy [29] result from the interaction of SSNOs, while coupled systems of SSNOs have been shown to have image processing capabilities [24,30] and have been proposed for associative memories [31,32]. In this paper, we first review recent work that establishes that SSNOs can also serve as substrates for general-purpose Boolean computation.…”