In our brain, close range interactions among electrically active neurons builds up a collective input-specific response to external stimuli able to trigger efficient cognitive tasks, like classification, decision-making and perception. In the same way, interacting nano-oscillators will be used to build a non-linear dynamical system capable of collective evolution, criticality and chaotic behavior as a basis for artificial computing machines and autonomous systems.
The goal of COSMO is to demonstrate, at the hardware and modeling levels, an analogue computational architecture based on interacting nanoscale nonlinear memristor-based oscillators. Hardware versions of memristors, i.e. metal/insulator/metal devices featuring electrically driven resistance change, together with scalable electronic components (resistors, capacitors) will constitute oscillators that will interact among each other within network architectures. The resulting collective dynamics will be exploited to design unconventional computational schemes for pattern storage and mapping of input signal according to associative memory.
The research will be conducted from material science, device physical and analytical modeling, to modeling of dynamical features of oscillator arrays and up to the elaboration of collective computational schemes. The ambition is to provide a progress in all the listed topic culminating in a radical paradigm shift in the information processing towards a non-Boolean, robust and efficient computing system.
- Assessment of computational primitives of memristor devices
- Provide new tools for memristor device modeling and circuit synthesis
- Provide a computational model for analogue computing with coupled memristor oscillators
CNR - Principal Investigator: Stefano Brivio
Involved Personelle: Sabina Spiga
Call for personelle: post-doc position