Time: Friday, October 12th, 11:40
Speaker: Giovanni FINOCCHIO, UNIME
The recent availability of big data and computational power (Graphic Processing Unit) lunched the application of deep learning, machine learning and artificial intelligence (AI) in real-life applications. Most of the solutions are software oriented and the time required for the setting proper parameters of the neural networks need a lot of time. Among all the possible hardware implementations, spintronics seems to be a candidate for that having the potential to work at low power and the possibility to exhibits strong non-linearity that is fundamental for the performance of AI systems.
In this talk, I will discuss recent results achieved in Messina in the development of spintronic devices that can be used for unconventional computing, development of AI, and neuromorphic applications. The first part of the talk will be oriented to the presentation of spin-Hall random bit generators and their possible impact in computing and security. Finally, I will discuss some possible configurations to use antiferromagnetic devices and skyrmions for unconventional computing and for approaching problem that can be converted in a topological network.