We focus on creating links to European and National projects and larger initiatives that support and develop different aspects of Neuromorphic Computing Technologies.
For this, we developed cartographical analysis deliverable which includes:
Although the document was intended to focus on current projects, we have added past projects (not older than 2 years) to involve as many people as possible within Neurotech community.
The aim of State of Art deliverable is to summarize the current status of all types of neuromorphic hardware technology from fully CMOS-based systems to solutions exploiting the use of memristive devices, advanced device concepts in the field of spintronic and photonics, and novel materials including 2D, nanowires and organic materials.
Section I introduces the current state of the art of large-scale neuromorphic computing systems based on digital CMOS or analogue/mixed-signal technologies.
Section II introduces the various types and physical mechanisms of memristive device technologies, which include a broad class of two or three-terminal devices whose resistance can be modified upon electrical stimuli. Moreover, we describe the proposed hardware implementation of synaptic and neuronal circuits exploiting those memristive technologies, which are currently at high maturity level, namely resistive random access memory (RRAM), phase change memory (PCM), ferroelectric memory (FeRAM), and magnetoresistive random access memories (MRAM), metal-insulator-transition (MIT) devices, as well as more explorative and innovative concepts.
Section III introduces the current state of the art of mixed CMOS-memristive device neuromorphic chips. While in the previous section we discuss mainly the implementation of some specific neuromorphic function by exploiting single or small blocks of memristive devices, in this section we summarize the current state of monolithic integrated CMOS-memristive devices in a chip, or of large systems demonstrated at mixed software – hardware level. Currently this section includes mainly the hybrid CMOS-RRAM neuromorphic chip, and IBM work on PCM, but it will be updated in the future with other technologies and future advancements.