Emergent spiking in non-ideal memristor networks

Abstract

Memristors have uses as artificial synapses and perform well in this role in simulations with artificial spiking neurons. Our experiments show that memristor networks natively spike and can exhibit emergent oscillations and bursting spikes. Networks of near-ideal memristors exhibit behaviour similar to a single memristor and combine in circuits like resistors do. Spiking is more likely when filamentary memristors are used or the circuits have a higher degree of compositional complexity (i.e. a larger number of anti-series or anti-parallel interactions). 3-memristor circuits with the same memristor polarity (low compositional complexity) are stabilised and do not show spiking behaviour. 3-memristor circuits with anti-series and/or anti-parallel compositions show richer and more complex dynamics than 2-memristor spiking circuits. We show that the complexity of these dynamics can be quantified by calculating (using partial auto-correlation functions) the minimum order auto-regression function that could fit it. We propose that these oscillations and spikes may be similar phenomena to brainwaves and neural spike trains and suggest that these behaviours can be used to perform neuromorphic computation. arXiv:1210.8024v2 [cond-mat.mtrl-sci]

Nanoarchitectonic atomic switch networks for unconventional computing

“…Under applied bias, the completion of a filament at one location causes a potential drop in another location, which in turn influences other connected junctions, resulting in the propagation of switching activity through internal feedback. 27) These processes constantly reconfigure the network and result in time-dependent potential maps of electrical activity recorded by the measurement electrodes where cascading local or distributed switching events are readily visualized. 19) Continuously evolving voltage traces are observed at points distributed throughout the network (Fig.…”

Neural Net to Neuronal Network Memristor Interconnects

“…can memristors act as an input to a living neuronal network) and if the spiking cells can alter the memristor dynamics -the latter question concerns us here. Various studies [15,16,20] have shown that the memristor spikes are reproducible and can interact through time on a single memristor and/or through space in a network of memristors. Here we show that neuronal cell spiking can affect the spiking properties of an attached memristor network.…”

“…Previous work [15,20,16] has drawn attention to the short-term memory aspects of the memristor, which might be useful in copying neuronal networks. The d.c. response of the memristor is a current spike that contains the short-term memory of the memristor [15]; on a single memristor these spikes can be used to compute in a Boolean fashion [16], and networks of memristors exhibit brainwave-like dynamics and bursting spikes [20].…”

“…The d.c. response of the memristor is a current spike that contains the short-term memory of the memristor [15]; on a single memristor these spikes can be used to compute in a Boolean fashion [16], and networks of memristors exhibit brainwave-like dynamics and bursting spikes [20]. The similarity between memristor network dynamics, and brainwaves and neuronal spiking patterns suggests that the memristor could be a good choice for either copying the brain (i.e.…”

Structural and Material Changes in Thin Film Chalcogenide Glasses Under Ar-Ion Irradiation

“…At the same time, these energies provide a well-established sputtering velocity, yielding defined RCBM device arrays with excellent device characteristics, such as uniform and switching behavior. Integrated circuits based solely on memristor arrays show a great potential for new circuit solutions as demonstrated by Cali et al [28].…”