A Quantized Convolutional Neural Network Implemented With Memristor for Image Denoising and Recognition

Zhang, Yuejun; Wu, Zhixin; Liu, Shuzhi; Guo, Zhecheng; Chen, Qilai; Gao, Pingqi; Wang, Pengjun; Liu, Gang

doi:10.3389/fnins.2021.717222

Cited by 6 publications

(4 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It not only dramatically enhances the data storage capability of gadgets, but it also allows neuromorphic systems to analyze information more efficiently. Multiple conductance states of memristor devices should, in general, be realized in as simple a manner as possible for practical use [ 45 – 47 ]. Nonetheless, the most investigations to-date have relied on sophisticated programming approaches to obtain multi-conductance characteristics, with the addition of varying current compliances and voltages putting a strain on the overall circuit design [ 30 , 48 , 49 ].…”

Section: Introductionmentioning

confidence: 99%

Robust Resistive Switching Constancy and Quantum Conductance in High-k Dielectric-Based Memristor for Neuromorphic Engineering

et al. 2022

View full text Add to dashboard Cite

For neuromorphic computing and high-density data storage memory, memristive devices have recently gained a lot of interest. So far, memristive devices have suffered from switching parameter instability, such as distortions in resistance values of low- and high-resistance states (LRSs and HRSs), dispersion in working voltage (set and reset voltages), and a small ratio of high and low resistance, among other issues. In this context, interface engineering is a critical technique for addressing the variation issues that obstruct the use of memristive devices. Herein, we engineered a high band gap, low Gibbs free energy Al2O3 interlayer between the HfO2 switching layer and the tantalum oxy-nitride electrode (TaN) bottom electrode to operate as an oxygen reservoir, increasing the resistance ratio between HRS and LRS and enabling multilayer data storage. The Pt/HfO2/Al2O3/TaN memristive device demonstrates analog bipolar resistive switching behavior with a potential ratio of HRS and LRS of > 105 and the ability to store multi-level data with consistent retention and uniformity. On set and reset voltages, statistical analysis is used; the mean values (µ) of set and reset voltages are determined to be − 2.7 V and + 1.9 V, respectively. There is a repeatable durability over DC 1000 cycles, 105 AC cycles, and a retention time of 104 s at room temperature. Quantum conductance was obtained by increasing the reset voltage with step of 0.005 V with delay time of 0.1 s. Memristive device has also displayed synaptic properties like as potentiation/depression and paired-pulse facilitation (PPF). Results show that engineering of interlayer is an effective approach to improve the uniformity, ratio of high and low resistance, and multiple conductance quantization states and paves the way for research into neuromorphic synapses.

show abstract

Section: Introductionmentioning

confidence: 99%

Robust Resistive Switching Constancy and Quantum Conductance in High-k Dielectric-Based Memristor for Neuromorphic Engineering

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Multi-state switching characteristics are essential for memristors to realize the full potential of neuromorphic computing functions. 43,44 Multiple resistance states can be obtained in mSiO 2 based memristors by controlling the CC. Fig.…”

Section: Resultsmentioning

confidence: 99%

3D-structured mesoporous silica memristors for neuromorphic switching and reservoir computing

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Finally, a bipolar memristor with asymmetric resistive switching characteristics was formed [ 7 , 9 ]. Figure 2 a,b shows the Pt/ZnO/Pt-structured memristor prepared in the laboratory, which first deposited a 50 nm thin film of ZnO as a resistive layer on a commercial Pt/Ti/SiO2 wafer using RF magnetron sputtering, and then deposited a 50 nm thick Pt electrode to form a Pt/ZnO/Pt structure using electron-beam deposition, where the specific fabrication process and device characteristics have been described in detail in a previous work [ 41 ]. Figure 2 c shows the transmission electron microscopy image of the memristor cross section.…”

Section: Methodsmentioning

confidence: 99%

Reservoir Computing-Based Design of ZnO Memristor-Type Digital Identification Circuits

et al. 2022

View full text Add to dashboard Cite

Reservoir Computing (RC) is a network architecture inspired by biological neural systems that maps time-dimensional input features to a high-dimensional space for computation. The key to hardware implementation of the RC system is whether sufficient reservoir states can be generated. In this paper, a laboratory-prepared zinc oxide (ZnO) memristor is reported and modeled. The device is found to have nonlinear dynamic responses and characteristics of simulating neurosynaptic long-term potentiation (LTP) and long-term depression (LTD). Based on this, a novel two-level RC structure based on the ZnO memristor is proposed. Novel synaptic encoding is used to maintain stress activity based on the characteristics of after-discharge and proneness to fatigue during synaptic transmission. This greatly alleviates the limitations of the self-attenuating characteristic reservoir of the duration and interval of the input signal. This makes the reservoir, in combination with a fully connected neural network, an ideal system for time series classification. The experimental results show that the recognition rate for the complete MNIST dataset is 95.08% when 35 neurons are present as hidden layers while achieving low training consumption.

show abstract

A Quantized Convolutional Neural Network Implemented With Memristor for Image Denoising and Recognition

Cited by 6 publications

References 41 publications

Robust Resistive Switching Constancy and Quantum Conductance in High-k Dielectric-Based Memristor for Neuromorphic Engineering

Robust Resistive Switching Constancy and Quantum Conductance in High-k Dielectric-Based Memristor for Neuromorphic Engineering

3D-structured mesoporous silica memristors for neuromorphic switching and reservoir computing

Reservoir Computing-Based Design of ZnO Memristor-Type Digital Identification Circuits

Contact Info

Product

Resources

About