Miguel Romera Rabasa is an experimental scientist specialising in Spintronics, Nanomagnetism and Bio-inspired Computing. He began his research career at the Polytechnic University, where he obtained his doctoral thesis in 2012 with the highest qualification cum laude and international mention, after three international stays in laboratories of recognised prestige. His post-doctoral stage takes place in France, where he works in world-leading research centres in the field of Spintronics. He is doing his first post-doctorate at Spintec (Grenoble), a laboratory that performs both basic research and the technological development of end devices. At Spintec, he acquired an applied vision of scientific research, which was reinforced during his second post-doctorate, at the CNRS associated unit with the leading technology company Thales (Paris). During this period, he came into contact with the field of Neuromorphic Computing, which together with Spintronics constitutes his current line of research. He was a ComFuturo researcher from 1 September 2018 to 31 January 2019 in the Madrid Institute of Micro and Nanotechnology where he began the development of a project titled “Artificial neural network based on spintronic nano-devices”.
In the era of Big Data, processing the increasing amount of information generated by our society on a daily basis is driving computing systems to enormous levels of energy consumption. Thus, information technologies already consume 10% of the energy produced in the world.
This problem, coupled with the new demands coming from the field of Artificial Intelligence, which require processors capable of performing cognitive tasks, has led hardware developers to seek inspiration from the most efficient processor known: the human brain, capable of classifying images in fractions of a second while consuming less energy than a light bulb.
The concept of Neuromorphic Computing (computing systems inspired by the human brain) is considered by the scientific community to be one of the most disruptive technologies with the greatest potential to transform data processing systems in the short-medium term. Thus, it is expected to give rise to intelligent and highly adaptive computing systems with very low energy cost, generating an immense impact in various fields such as Big Data or Artificial Intelligence.
Developing an artificial neural network in hardware requires the parallel connection of complex processing units (neuron-like) coupled by adjustable connections that implement memory (synapse-like). In this context, advances in nanotechnology and spin electronics (Spintronics) have enabled the development of new magnetic nano-devices that offer a unique opportunity to mimic the behaviour of neurons and synapses at the “nano” scale with low energy cost.
This project aims to develop the proof of concept of a neuromorphic computing system based on magnetic nano-devices already implemented in industrial applications, capable of performing cognitive tasks and with learning capabilities, while maintaining a minimum energy consumption.
Application: The technology proposed in this project aims to lead to the development of energy-efficient neural networks on chip. This would have an enormous impact on different fields such as Artificial Intelligence, Big Data or the microelectronics industry. There are many potential applications: virtual assistants for smartphones, autonomous cars, internet search engines, integrated systems for automatic management of Big Data or biomedical prostheses.