H2020Индивидуална стипендия2018–2020

NeuroMag · Magnonic Matrix-Vector-Multiplier for Neural Network Applications

„Хоризонт 2020“ — Действия „Мария Склодовска-Кюри“

Период: 2018-05-22 → 2020-05-21
Финансиране от ЕС: 172 800 €
Участници: 1
Схема: MSCA-IF-EF-ST

Линиите свързват координатора с партньорите.

Накратко на български

Спин-вълните в специални магнитни филми се използват за извършване на математически операции, като умножението на матрица по вектор. Това помага за повишаване на ефективността при обработката на сигнали в изкуствените невронни мрежи.

Този кратък обзор е генериран от изкуствен интелект

Кратко обяснение, генерирано от езиков модел по текста на CORDIS. Оригиналът е по-долу.

Резултати накратко

Magnonic Matrix-Vector-Multiplier for Neural Network Applications

The most computationally intensive part in a neural network (both in training and operation) are the matrix-vector multiplications (large-scale linear transformations) needed to propagate signals between the different layers. To improve the efficiency of matrix vector multipliers (MVMs), it has been proposed to perform the linear transformations through physical processes instead of digital computation. Key properties of a matrix-vector multiplier (MVM) in artificial neural networks are its bidirectionality (ability to perform the multiplication with the transposed matrix in the opposite direction) and trainability (ability to adapt itself to the forward and the error signal). Trainable bidirectional MVMs have been suggested based on the interference of waves. The implementation of NeuroMag explored and laid the groundwork towards MVMs using the propagation, interference and phase shift (training) of spin waves. High quality epitaxial Y3Fe5O12 (Yttrium Iron Garnet, YIG) films were developed as a medium for spin waves. The low magnetic damping in YIG limits detrimental effects due to spin wave attenuation. Electrical to magnetic transducers can act as generators, detectors and training mechanism, leading to full bidirectionality and trainability of the MVM. To reach this ambitious goal, over the course of 2 years, starting from May 22, 2018 and ending on May 21, 2020, the researcher targeted 3 main research objectives: 1. To achieve a pulsed laser deposition process for epitaxial YIG thin films on Ga3Gd5O12 substrates as ferromagnetic medium for the propagation and interference of spin waves. 2. Develop magnetoelectric transducers to generate and detect spin waves. 3. Develop a magnonic MVMs and demonstrate its operation. To meet these goals, each objective was addressed to a distinct work package (WP1-3) with tangible targets and specifications.

Текст от CORDIS, на английски · Данни: CORDIS, © Европейски съюз

Цел на проекта

Machine learning applications based on artificial neural networks have undergone rapid progress in recent years. To improve the power efficiency over current hardware, alternative implementations of a critical part of artificial neural networks, the matrix-vector multiplier performing large-scale linear transformations, have been intensively researched. Recently, a matrix-vector multiplier based on the interference of optical waves has been proposed in combination with local adjustable electro-optic modulation of the refractive index to enable training. NeuroMag’s objective is to implement such an interference-based matrix-vector multiplier using spin waves (magnons). Magnetoelectric compound materials will be used to engineer scalable broadband transducers with high potential energy efficiency to generate, detect, and manipulate spin waves. Distinct advantages of such a spin wave implementation over a photonic one are (i) the full bidirectionality of the system since transducers can be operated both to excite as well as detect spin waves and (ii) the large tuning range of the phase velocity of spin waves (equivalent to the refractive index in photonics) by the magnetoelectric effect. Magnetoelectric transducers and low-damping Yttrium Iron Garnet magnetic media will be combined to nanofabricate a demonstrator device and study its matrix-vector multiplier operation.Using an interdisciplinary approach that combines materials science, physics, microwave engineering, and device nanofabrication, NeuroMag thus targets the ground-breaking proof-of-concept of a magnonic matrix-vector multiplier and its operation, paving the way towards magnonic artificial neural networks. The combination of learning through research and a comprehensive training plan, including both scientific and technological as well as soft skills, will strongly enhance the researcher profile of the applicant and provide a boost for his future scientific career.

Оригинален текст от CORDIS (на английски).

Участници

INTERUNIVERSITAIR MICRO-ELECTRONICA CENTRUM · LeuvenКоординаторБелгия