Descripción del proyecto
Computación neuromórfica con «neuronas» capaces de oscilar de forma sincronizada
La computación neuromórfica utiliza métodos basados en «hardware» y «software» para simular la estructura y el funcionamiento del cerebro humano. La inteligencia artificial es un campo bien desarrollado, pero la emulación de neuronas y sinapsis con «hardware» es un ámbito de aplicación emergente. Las oscilaciones eléctricas son un fenómeno común en el cerebro vinculado a diversos procesos cognitivos. El equipo del proyecto SMART, que cuenta con el apoyo de las Acciones Marie Skłodowska-Curie, aprovechará la resistencia diferencial negativa (NDR, por sus siglas en inglés), una propiedad de algunos componentes que les permite producir oscilaciones en respuesta a picos eléctricos aplicados, para emular el comportamiento de las neuronas biológicas. Se integrarán películas delgadas de un nuevo material como prueba de la NDR en redes neuronales oscilantes para demostrar tareas informáticas complejas.
Objetivo
Rapid progress in the regime of Artificial Intelligence and Internet-of-Things has enthused the development of fast and energy-efficient hardware to support future computing needs. One of the most prominent solutions is the deployment of cross-disciplinary resources for mimicking the performance of a Human Brain, also known as neuromorphic computing. It relies on electronic components that could replicate the functioning of neurons and synapses. Designing such novel electronics needs the development of cost-effective, fast and reliable materials with tunable functionality. A fundamental understanding of these materials will pave the foundation of innovative device designs and strategies for their large-scale integration for neuromorphic architectures. In this context, the proposed project intends to deliver artificial neurons by developing thin-films of a novel material (TbMnO3) that is capable of demonstrating a Negative Differential Resistance (NDR). The idea is to utilise the fundamental understanding of multi-dimensional (electrical, optical, mechanical and magnetic) control of NDR which is not explored yet and is possible in TbMnO3. NDR enables a two-terminal device to display self-oscillations due to applied electrical spikes. The spiking electrical currents could be used to govern the behaviour of these oscillations and emulating the leaky, integrate, and fire behaviour of biological neurons. Once such a performance is achieved, the device will further be integrated into oscillatory neural networks array for demonstration of complex computing tasks such as image recognition and imitating human behaviour.
The project will unite the applicant’s expertise in synaptic devices and materials engineering with the extensive experience of the host-labs in thin-film and neuromorphic devices. Importantly, the project will warranty catapulting the applicant’s international recognition as an independent researcher and will improve his career prospects.
Ámbito científico
Programa(s)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Régimen de financiación
HORIZON-AG-UN - HORIZON Unit GrantCoordinador
9712CP Groningen
Países Bajos