Abstract
Multiple modern applications of electronics call for inexpensive chips that can perform complex operations on natural data with limited energy. A vision for accomplishing this is implementing hardware neural networks, which fuse computation and memory, with low cost organic electronics. A challenge, however, is the implementation of synapses (analog memories) composed of such materials. In this work, we introduce robust, fastly programmable, nonvolatile organic memristive nanodevices based on electrografted redox complexes that implement synapses thanks to a wide range of accessible intermediate conductivity states. We demonstrate experimentally an elementary neural network, capable of learning functions, which combines four pairs of organic memristors as synapses and conventional electronics as neurons. Our architecture is highly resilient to issues caused by imperfect devices. It tolerates inter-device variability and an adaptable learning rule offers immunity against asymmetries in device switching. Highly compliant with conventional fabrication processes, the system can be extended to larger computing systems capable of complex cognitive tasks, as demonstrated in complementary simulations.
Highlights
Biology-inspired electronics is currently attracting increasing attention as modern applications of electronics, such as biomedical systems, ubiquitous sensing, or the future Internet-of-Things, require systems able to deal with significant volumes of data, with a limited power budget
Each row of the system is referred to as a Single Neuron Unit (SNU); all rows for a given layer are connected to a finite state machine (FSM) which provides programming and input learning pulses[34]
The circuit is composed of a CMOS-based neuron programmed by a field-programmable gate array (FPGA), and a series of memristive devices mimicking synapses between differential inputs and the neuron
Summary
Biology-inspired electronics is currently attracting increasing attention as modern applications of electronics, such as biomedical systems, ubiquitous sensing, or the future Internet-of-Things, require systems able to deal with significant volumes of data, with a limited power budget. Memristive devices can be made with organic materials that are fundamentally attractive[22,23] as they offer unique advantages: low material costs, scalable fabrication via roll-to-roll imprint lithography, and compatibility with flexible substrates These properties pave the way towards integration with embedded sensors, bio-medical devices, and other internet of things applications[24,25], yet often come at the cost of slower programming relative to inorganic memristive devices or binary organic memory devices[26,27]. An under-explored topic in moving from device simulations to real hardware prototypes is the intrinsic variability of memristive devices Other imperfect behaviors, such as an asymmetric increase (SET) and decrease (RESET) of device conductance in filamentary-based memristors, resistance instability in phase change devices, and stuck-on/off effects complicate deterministic learning strategies even further[29,30]. These findings suggest ways to improve our device, but neuromorphic supervised learning systems in general
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
