Artificial intelligence has been object of research for the last 60 years and neural networks empowered by deep learning algorithms are now able to deal with very difficult tasks such as translate sentences, play complex games (i.e. Go) and pass touring test. All these tasks were implemented in software and run on high performance processors in standard CMOS technology. However, scaling these systems up to the complexity of human brain is not trivial due to the limits of von-Neumann architecture (physical separation of memory and computational unit) and device scaling, indeed Moore’s law, which led the electronic industry since the 60’s, is approaching its end. Thus, new architectures and devices are needed to investigate the scalability of neural networks to the complexity of the brain. In this framework, neuromorphic architectures, where the synapses-neurons connections and the weight update algorithms are inspired by the biological world, are gaining interest among the academy and industries.
My research activity is focused on the design and hardware implementation of neuromorphic networks with novel memristive device.