Dresden 2017 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
BP: Fachverband Biologische Physik
BP 42: Neurosciences
BP 42.2: Vortrag
Mittwoch, 22. März 2017, 15:30–15:45, ZEU 250
Hippocample learning with memristive devices: device requirements for the use in recurrent networks — •Nick Diederich1,2, Thorsten Bartsch2, Martin Ziegler1, and Hermann Kohlstedt1 — 1Technische Fakultät, Christian-Albrechts-Universität zu Kiel — 2Neurologie, Universitätsklinik Schleswig-Holstein
Memristive devices are considered as promising candidates for hardware based synapses since they fulfill important biological plasticity rules such as long-term potentiation and long-term depression. Furthermore, their low energy consumption, scalability, and rather simple device structure are especially interesting for artificial neural networks. In this talk, the opportunities of memristive devices for recurrent neural-networks are presented. Those networks are in particular important structures in mammal brains. In detail, a memristive network-model of the hippocampale loop is presented which allows the realization of physiological behavior of learning. The unique behaviors and desired device characteristics of memristive devices for those network structures will be discussed.
Financial support by the German Research Foundation through FOR 2093 is gratefully acknowledged.