Description du projet
Exploiter les matériaux topologiques dans une architecture de réseau neuronal plus «semblable au cerveau»
Bien que le cerveau ressemble à une boule de spaghettis, il est composé d’environ 86 milliards de cellules nerveuses formant quelque 100 000 milliards d’interconnexions. Par ailleurs, ces interconnexions sont très ordonnées, de sorte que différentes régions du cerveau remplissent différentes fonctions. La réalisation de ce type d’interconnexion s’est avérée particulièrement difficile dans les architectures de réseaux neuronaux physiques actuelles, issues de l’ingénierie neuromorphique. Le projet SCHINES, financé par l’UE, relève ce défi en réunissant des chercheurs de renommée mondiale pour mettre au point un nouveau type d’architecture dotée d’une interconnectivité évolutive et exploitant les propriétés exotiques des matériaux topologiques.
Objectif
Creating a brain-inspired technology through neuromorphic engineering could achieve or even surpass the extraordinary ability of the brain to grasp the world, which operates at an extremely low power consumption yet with the most complex interconnectivity known to mankind. The main goal of SCHINES is to set a clear direction to solve one of the biggest technological challenges that hinders this revolution: in existing physical neural network architectures, the desired interconnectivity can hardly be achieved. We will fabricate and design devices to demonstrate radically improved signal routing using topological metals. The design principle is simple: the environment of chiral electrons, electrons with spin locked to its momentum, can be engineered to create rich electronic lensing effect, analogous yet broader to light in-media propagation. Positive and negative effective indices of refraction for electrons, and lossless signal crossing can be engineered while maintaining, selecting or filtering the intrinsic topological protection of chirality, a degree of freedom that can be used for computation. These design principles are the basis for our device goal with scalable interconnectivity and are highly transferrable: they apply to strained materials, magnetic domains and heterostructures. This ambitious goal is realistic due to the interdisciplinary breadth of the SCHINES consortium: it is built out of established and emerging leaders called to shape the future of the field, joined in a public-private collaboration. They comprise an expertise that bridges the gap between the most abstract quantum field theory calculations with microscopic modelling with sample fabrication and measurement finalizing in device assembly
Champ scientifique
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringsignal processing
- natural sciencesphysical sciencesquantum physicsquantum field theory
- natural sciencesbiological sciencesneurobiologycomputational neuroscience
- natural sciencescomputer and information sciencesartificial intelligencecomputational intelligence
Mots‑clés
Programme(s)
Régime de financement
RIA - Research and Innovation actionCoordinateur
8803 Rueschlikon
Suisse