The transversal team aims at bringing together all competencies from SPINTEC involving spintronic devices nanofabrication, characterization, circuit integration, architecture and algorithm techniques to implement hardware solutions for AI and unconventional computing applications that are hard to be with current CMOS technologies.
Spintronic based multifunctional devices are also a great opportunity to improve the energy efficiency. Moreover, the high speed and endurance magnetic devices allow more sophisticated learning algorithms implementations such as unsupervised learning, reinforcement learning, etc, crossing the gap between current edge AI implementations and the real brain computing ability.
As the brain performs very sophisticated operations and consumes only few Watts, brain-inspired/neuromorphic computing is very promising path for which spintronic devices can efficiently emulate both neurons and synapses in hardware. Their nanometric size, sensitivity to input stimuli, and interactions makes those devices ideal to implement large arrays of neuro-synaptic elements. In addition, these types of architectures are very well suited to new learning rules and adaptation. Performant and highly energy efficient, analog, voltage or current-tunable spintronic-based memristors can emulate neurons and massive dynamic connections required to solve complex, non-linear operations: Spintronic nanoscillators, spintronic and ferroelectric memristors, magnetic memories, super paramagnetic and magnetic tunnel junctions, skyrmions, etc.
Biologically inspired operations can take huge benefit from noise, which, in current nanotechnology is considered as an issue. Indeed, noise can be a crucial ingredient to emulate the stochastic nature of neural activity and used for computing. In this context, probabilistic computing is particularly suitable approach that relaxes usual precision constraints. Stochastic circuits and architectures largely benefit from spintronic based device implementations aiming at performing new energy-based or temporal-based machine learning models. The truly random nature of spintronic devices (such as magnetic and super paramagnetic tunnel junctions) makes them attractive for hardware implementations of AI and unconventional computing.
In memory computing
The most promising solutions for non-Von Neumann, in-memory computing architectures are based on the use of emerging technologies, that are able to act as both storage and information processing units thanks to their specific physical properties. High accuracy, Deep neural networks (DNN) can be built with crossbars analog in memory computing concept, involving MRAM families, such as STT, SOT, VCMA, but also with more exotic families of magneto-resistive, and ferroelectric or skyrmion based devices.
- IRS SPINBRAIN (2020-2022)
- ANR Spinspike (2020-2024)
- HANS, UGA (2019-2022)
- CEA LIST
- CEA LETI
- Seminar – Chiral magnetism controlled by spin-orbit torques and spin-torque nano-oscillators: From magnetic memories to neuromorphic computing (May 29th, 2019)
On Monday June 24 at 14:00 we have the pleasure to welcome Jacob Torrejon from SPEC, CEA Saclay, France. He will give us a seminar at CEA/IRIG, Bat 1005, room 445 entitled : Chiral magnetism controlled ...
- Masters thesis projects for Spring 2019 (October 10th, 2018)
You find here the list of proposals for Master-2 internships to take place during Spring 2019. In most cases, these internships are intended to be suitable for a longer-term PhD work. Interested Master-1 students are ...
- Master students to visit SPINTEC and discuss our topics for internships (September 18th, 2016)
On 25th October 2016 our host Institut INAC welcomes students for a presentation of internship topics proposed to host Master-2 students during Spring 2017. Details will be provided later.