Associative memory by virtual oscillator network based on single spin-torque oscillator

Abstract

A coupled oscillator network may be able to perform an energy-efficient associative memory operation. However, its realization has been difficult because inhomogeneities unavoidably arise among the oscillators during fabrication and lead to an unreliable operation. This issue could be resolved if the oscillator network were able to be formed from a single oscillator. Here, we performed numerical simulations and theoretical analyses on an associative memory operation that uses a virtual oscillator network based on a spin-torque oscillator. The virtual network combines the concept of coupled oscillators with that of feedforward neural networks. Numerical experiments demonstrate successful associations of 60-pixel patterns with various memorized patterns. Moreover, the origin of the associative memory is shown to be forced synchronization driven by feedforward input, where phase differences among oscillators are fixed and correspond to the colors of the pixels in the pattern.

Spintronic virtual neural network by a voltage controlled ferromagnet for associative memory

“…If we can give a correspondence between the sign of and, for example, black-and-while colors in a pattern, an associative memory operation will be possible by manipulating the VCMA device. For example, in the associative memory operation of black-and-white color patterns using an oscillator network 20 , 24 , ( ), or equivalently ( ), means that the color of the i th pixel is white (black). Bearing this in mind, in this work, we define the correspondence between the color and the magnetization state by parameters as where .…”

“… 38 . Thus, one might consider to replace an STO in the virtual oscillator network 22 , 24 with this parametric oscillator and develop an oscillator network. However, the fact that an external magnetic field is necessary for this parametric oscillator 38 suggests that such method is unsuitable for practical application.…”

“…Now let us explain how to perform the associative memory operation by a single VCMA device. Similar to the virtual oscillator network by an STO 24 , the associative memory operation by the virtual network consists of three steps. Each step consists of applying voltage N times, and therefore, totally the voltages should be applied 3 N times.…”

“…The present system, however, consists of a single device, which is the different aspect from the conventional neural networks. In addition, the present system is also different from the virtual oscillator network 24 . This is because the algorithm developed here requires the relaxed state of the magnetization only for the computation, while the virtual oscillator network requires long-time memories of the phase during the magnetization oscillation in an STO.…”

“…As revealed in Ref. 24 , the operation principle of the virtual oscillator network is similar to a feedforward neural network, rather than the conventional, instantaneously coupled oscillator networks 16 – 20 . The virtual oscillator network solved several issues in the conventional oscillator networks 22 , such as unstable operation due to inhomogeneity in the oscillators.…”