Controllable Multiple Depression in a Graphene Oxide Artificial Synapse

Yang

et al. 2019

Small

The inherent limitations of traditional silicon technology in the implementation of machine learning could be addressed by neuromorphic computing since it is able to mimic the resilience, versatility, and efficiency of the human brain. [1][2][3][4] Synaptic It is desirable to imitate synaptic functionality to break through the memory wall in traditional von Neumann architecture. Modulating heterosynaptic plasticity between pre-and postneurons by another modulatory interneuron ensures the computing system to display more complicated functions. Optoelectronic devices facilitate the inspiration for high-performance artificial heterosynaptic systems. Nevertheless, the utilization of near-infrared (NIR) irradiation to act as a modulatory terminal for heterosynaptic plasticity emulation has not yet been realized. Here, an NIR resistive random access memory (RRAM) is reported, based on quasiplane MoSe 2 /Bi 2 Se 3 heterostructure in which the anomalous NIR threshold switching and NIR reset operation are realized. Furthermore, it is shown that such an NIR irradiation can be employed as a modulatory terminal to emulate heterosynaptic plasticity. The reconfigurable 2D image recognition is also demonstrated by an RRAM crossbar array. NIR annihilation effect in quasiplane MoSe 2 /Bi 2 Se 3

mentioning

confidence: 99%

Near‐Infrared Annihilation of Conductive Filaments in Quasiplane MoSe₂/Bi₂Se₃ Nanosheets for Mimicking Heterosynaptic Plasticity

Yang

et al. 2019

Small

“…Besides, the development of electronics enlightens us that, for a fundamental electronic element, the extension of additional function is usually able to advance the upper‐level functions in an integrated configuration. Otherwise the neglect of a certain property may induce the ineffectiveness of an electronic chip . Based on this consideration, a new path is paved with a view to the neuromorphic computing to construct the new type artificial synapse utilizing this specific memristor and emulate the fresh neuroplasticity for novel bioinspired computing .…”

mentioning

confidence: 99%

Rectification‐Regulated Memristive Characteristics in Electron‐Type CuPc‐Based Element for Electrical Synapse

Yang

Zhu

et al. 2017

Adv Elect Materials

Self Cite

As one novel electronic strategy, the integration of memristive behaviors with rectification has shown its potential application. Herein, based on the interfacial effects of a LiF buffer layer at the anode of an electron‐type CuPc memristor on reconfiguring the device energy levels and redistributing the charge accumulations, a new class of organic memristor with rectifying characteristics is proposed. Memristive behaviors characterized by smooth consecutive changes are obtained, which are illustrated in detail by the adjustable device energy levels and relevant carrier processes. For this device, aside from its appeal in solving the well‐known cross‐talk problem, based on the consecutive rectification‐regulated memristive behaviors, this study paves the way for devising a new type artificial synapse—an electrical rectifying synapse for effective comprehensive bioinspired computations. Thus these results can invite further investigations of the memristor through energy‐level reconfiguration and open original perspectives for rectification‐regulated memristors in design for novel brain‐like information processing.

“…In this case, multiple depression plays a profound role in filtering the neural signals and regulating the overall neural functions. Given its irreplaceable functions, multiple depression deserves more attention . In addition, the commonly used rules are simplified rules that are missing various essential and sophisticated processes presented within the framework of realistic neural networks.…”

Section: Discussionmentioning

confidence: 99%

Nanoionics‐Enabled Memristive Devices: Strategies and Materials for Neuromorphic Applications

Nagai

et al. 2017

Adv Elect Materials

Self Cite

181

124

Memristors have been intensively studied in recent years as potential building blocks for the construction of versatile neuromorphic architectures. The prevalent developments focus on nanoionic devices due to their ease of reversible regulation, non-volatility, scalability down to the atomic level, low-energy consumption, nanosecond response times, sufficient yield and reliability, as well as their homologous regulatory mechanisms with biological synapses. However, despite their desirable structures and excellent properties, the previously reported properties of nanoionic devices are rather diverse and dissatisfactory for the delivery of reliable analog properties, which is the precondition for the imitation of brain-like functions. Here, the general requirements of neuromorphic engineering in terms of device structure, characterization parameters, synaptic functions and device engineering are introduced. A critical overview of the proposed nanoionic mechanisms for memristive switching is given, focusing particularly on providing fundamental insights into the strategies for regulating the adaptive memristive characteristics of devices that resemble the behaviors of biological synapses, which is an element of neural networks. In addition, the research progress in active materials (e.g., oxides, polymers, small molecules, bio-macromolecules and dopants), especially regarding the correlation among materials, is elaborated and a rational approach to provide insight into new design strategies is considered, and how to understand the memristor mechanisms and further achieve excellent memristive devices is discussed. Finally, the current challenges and several possible future research directions in this area are discussed.