Nonlinear random projection machines are efficient neural networks capable of classifying real-life data with lower computational demands compared to standard artificial neural networks. They are well-suited for hardware implementation using nonlinear devices, enabling the creation of low-power hardware neural networks.
We implement such a network using vortex-based spin-torque oscillators (STVOs), magnetic tunnel junctions (MTJs) that transform input signals nonlinearly at low power. We identify three physical parameters affecting the STVO dynamics and the network's performance during data classification. We demonstrate their impact on a simplified nonlinear separation task and optimize them using ultrafast data-driven simulations for image recognition on the MNIST dataset.
This approach holds potential for further hyperparameter optimization in STVO-based hardware random projection machines, and for the efficient development of custom neural architectures tailored for neuromorphic data classification.
Spintronic oscillators have gained important interest in recent years. However, a both fast and accurate model describing their dynamics is currently lacking. Here we propose an unconventional semi-analytical model capable of predicting the fundamental properties of the oscillators, when subjected to a spin-polarized current. Using data-driven corrections, the steady-state and transient regimes of oscillations can be simulated quantitatively. This data-driven model is more than two billion times faster than micromagnetic simulations. Such method paves the way for high-throughput simulation campaigns in any prospective applications, such as neuromorphic spintronics.
Over the past decades, artificial intelligence (AI) has made significant technological advances with the prospect of increased computer capabilities (e.g., automation in decision making and data processing) and acquired an increasingly important role in our everyday technological environment (Dall-E, ChatGPT, etc.). The main issue is that the digital silicon-based computing technologies are very energy-intensive while solving cognitive tasks such as speech or image recognition. We propose to tackle this issue by combining condensed matter physics (spintronics) and artificial intelligence to design nanoscale neuromorphic computing hardware to solve machine learning tasks.
A semi-analytical method (data-driven model) is used to predict the dynamics of a Spin-Torque Vortex Oscillator (STVO). This model relies on an improved analytical model based on the Thiele equation approach and micromagnetic simulations. The improved analytical model shows that the Ampère-Oersted field cannot be neglected and it describes quantitatively the STVO dynamics only in the resonant regime when the data-driven model allows to describe it in the steady-state oscillating regime as well. In addition, the model is 2.1 million time faster than simulations. It can be used to simulate the spin-diode effect and functionalize the STVOs for neuromorphic applications.
We report on microwave oscillations induced by spin-transfer-torque in metallic spin-valves obtained by electrodeposition
of Co-Cu-Co trilayer structures in nanoporous alumina templates. Using micromagnetic calculations
performed on similar spin-valve structures it was possible to identify the magnetization dynamics associated
with the experimentally determined microwave emission. Furthermore it appears that in our particular geometry
the microwave emission is generated by the vortex gyrotropic motion which occurs in, at least, one of the two
magnetic layers of our spin-valve structures. Microwave emission was obtained in the absence of any external
magnetic field with the appropriate magnetization configuration.
Conference Committee Involvement (2)
Spintronics XVIII
3 August 2025 | San Diego, California, United States
Spintronics XVII
18 August 2024 | San Diego, California, United States
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.