Berlin 2018 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

DS: Fachverband Dünne Schichten

DS 6: Thin Film Applications

DS 6.6: Vortrag

Montag, 12. März 2018, 16:15–16:30, H 0111

Unsupervised Hebbian learning experimentally realized with analogue memristive crossbar arrays — •Finn Zahari, Mirko Hansen, Hermann Kohlstedt, and Martin Ziegler — Chair of Nanoelectronics, Faculty for Electrical Engineering and Information Technology, Kiel University, Germany

Memristive devices are promising candidates to emulate synaptic behaviour in neuromorphic circuits in an efficient manner. Even though in the last couple of years a variety of materials and device structures were employed to fabricate memristive devices, there is still a gap between promising computing schemes and their hardware realization with memristive devices. We show that so called double barrier memristive devices can be integrated into crossbar architectures without the need for additional selector devices. These ionic memristive devices show a non-filamentary interface-based resistive switching behaviour with a high I-V nonlinearity and asymmetry as well as self-rectifying and self-limitation characteristics. They are used to realize selector-device-free 16x16 crossbar-arrays with 256 memristive devices. A local Hebbian learning scheme was utilized to perform unsupervised learning of visual patterns to demonstrate the applicability of the selector-free crossbars within a mixed signal circuit consisting of double barrier memristive devices as hardware synapses und software neurons.

Financial support by the German Research Foundation through FOR 2093 is gratefully acknowledged.