Time and rate dependent synaptic learning in neuro-mimicking resistive memories

Abstract

Memristors have demonstrated immense potential as building blocks in future adaptive neuromorphic architectures. Recently, there has been focus on emulating specific synaptic functions of the mammalian nervous system by either tailoring the functional oxides or engineering the external programming hardware. However, high device-to-device variability in memristors induced by the electroforming process and complicated programming hardware are among the key challenges that hinder achieving biomimetic neuromorphic networks. Here, a simple hybrid complementary metal oxide semiconductor (CMOS)-memristor approach is reported to implement different synaptic learning rules by utilizing a CMOS-compatible memristor based on oxygen-deficient SrTiO3-x (STOx). The potential of such hybrid CMOS-memristor approach is demonstrated by successfully imitating time-dependent (pair and triplet spike-time-dependent-plasticity) and rate-dependent (Bienenstosk-Cooper-Munro) synaptic learning rules. Experimental results are benchmarked against in-vitro measurements from hippocampal and visual cortices with good agreement. The scalability of synaptic devices and their programming through a CMOS drive circuitry elaborates the potential of such an approach in realizing adaptive neuromorphic computation and networks.

TiO2 Nanowire Array Memristive Devices Emulating Functionalities of Biological Synapses

“…The postsynaptic signal is therefore dependent on the action potential amplitude and frequency, called the spike‐rate and ‐amplitude dependent plasticity. [ 8–10 ] High frequencies or large stimulus amplitudes can lead to a persistent durable structural and functional change of the synapse, so that even after a long time, a similar action potential results in a higher postsynaptic potential than for the first stimulus. [ 11,12 ] Memristors have found to be able to emulate those synaptic functions and pave the way for neuromorphic computing.…”

Artificial Somatosensors: Feedback Receptors for Electronic Skins

“…Moreover, to acquire all the four states of nociceptor (threshold, relaxation, allodynia, and hyperalgesia), multiple CMOS circuit units are required. [ 21 ] Memristors, which mimic characteristics of human nervous system, [ 22 ] can essentially resolve the bottleneck due to their exceptional switching performance in sub‐nanometer scale. [ 15,23 ] Therefore, it is of great scientific and technological importance to develop a somatosensory, which responds against real‐life stimuli in the form of pressure, temperature, and pain, exploiting memristor as the fundamental unit.…”

Toward a generalized Bienenstock-Cooper-Munro rule for spatiotemporal learning via triplet-STDP in memristive devices

“…There are two types of triplet-STDP in neuroscience: the first-spike-dominating model and last-spike-dominating model proposed by Froemke et al and Wang et al, respectively 41,42 . Progress has been made in emulating these two types of triplet-STDP using first-order and second-order memristors 36,[43][44][45] . However, the generalization from triplet-STDP to the BCM learning rule has not yet been experimentally demonstrated in memristors.…”