A memristive deep belief neural network based on silicon synapses

Wang, Wei; Danial, Loai; Li, Yang; Herbelin, Eric; Pikhay, Evgeny; Roizin, Yakov; Hoffer, Barak; Wang, Zhongrui; Kvatinsky, Shahar

doi:10.1038/s41928-022-00878-9

“…In addition, Wang et al reported an important method for fabricating memristors based on a CMOS integrated circuit technology. 50 The schematic of memristive synapses based on two-terminal floating-gate devices is shown in Figure 6c, in which it can be found that the as-fabricated silicon-based neural synaptic devices show significantly high performance, which meets the requirements for neuromorphic computing. Further, they have implemented an ultrahigh-accuracy image recognition using the as-fabricated memristor device array, which is expected to be applied to intelligent recognition systems in AI.…”

Section: Memristor-based Chipsmentioning

confidence: 99%

Memristor-Based Artificial Chips

Sun,

Chen,

Zhou

et al. 2023

View full text Add to dashboard Cite

Memristors, promising nanoelectronic devices with in-memory resistive switching behavior that is assembled with a physically integrated core processing unit (CPU) and memory unit and even possesses highly possible multistate electrical behavior, could avoid the von Neumann bottleneck of traditional computing devices and show a highly efficient ability of parallel computation and high information storage. These advantages position them as potential candidates for future data-centric computing requirements and add remarkable vigor to the research of next-generation artificial intelligence (AI) systems, particularly those that involve brain-like intelligence applications. This work provides an overview of the evolution of memristor-based devices, from their initial use in creating artificial synapses and neural networks to their application in developing advanced AI systems and brain-like chips. It offers a broad perspective of the key device primitives enabling their special applications from the view of materials, nanostructure, and mechanism models. We highlight these demonstrations of memristor-based nanoelectronic devices that have potential for use in the field of brain-like AI, point out the existing challenges of memristor-based nanodevices toward brain-like chips, and propose the guiding principle and promising outlook for future device promotion and system optimization in the biomedical AI field.

show abstract

“…First, we divide the collected device structure parameters as well as the impact ionization coefficients into two parts: the training data and the test data. The DNN and the training data are then used to train a model for the impact ionization coefficients, and, finally, the test data are used to test the accuracy of the prediction model [ 19 , 20 , 21 , 22 ]. As shown in Figure 8 , the DNN used in this work consisted of four hidden layers, an input layer, and, finally, an impact ionization coefficient unit.…”

Section: Deep Neural Network For Impact Ionization Coefficient Predic...mentioning

confidence: 99%

Impact Ionization Coefficient Prediction of a Lateral Power Device Using Deep Neural Network

Cui

¹

,

Ma

²

,

Shi

³

et al. 2023

Micromachines

0

View full text Add to dashboard Cite

Nowadays, the impact ionization coefficient in the avalanche breakdown theory is obtained using 1-D PN junctions or SBDs, and is considered to be a constant determined by the material itself only. In this paper, the impact ionization coefficient of silicon in a 2D lateral power device is found to be no longer a constant, but instead a function of the 2D coupling effects. The impact ionization coefficient of silicon that considers the 2D depletion effects in real-world devices is proposed and extracted in this paper. The extracted impact ionization coefficient indicates that the conventional empirical impact ionization in the Fulop equation is not suitable for the analysis of 2D lateral power devices. The veracity of the proposed impact ionization coefficient is validated by the simulations obtained from TCAD tools. Considering the complexity of direct modeling, a new prediction method using deep neural networks is proposed. The prediction method demonstrates 97.67% accuracy for breakdown location prediction and less than 6% average error for the impact ionization coefficient prediction compared with the TCAD simulation.

show abstract

“…Memristors, which are two-terminal devices that change their electrical conductance and store analog values, have been widely employed as synapses in artificial neural networks due to their simple structure, non-volatile analog memory properties, and inherent ability to process vector-matrix multiplication when they are fabricated in a crossbar array structure. [8][9][10][11][12][13][14][15][16] Based on the desirable characteristics of the memristor crossbar array, various studies have demonstrated that the energy efficiency, hardware size, and inference speed can be effectively improved compared with the conventional CMOS-based processors 12,17,18 Despite the effectiveness of memristors as synapses in neural network systems, it is hard to train the weights of the neural network in a memristor crossbar array. Two-terminal memristors show long retention and fast speed, but they usually undergo severe non-linear and asymmetric update properties.…”

Section: Introductionmentioning

confidence: 99%

“…Memristors, which are two-terminal devices that change their electrical conductance and store analog values, have been widely employed as synapses in artificial neural networks due to their simple structure, non-volatile analog memory properties, and inherent ability to process vector-matrix multiplication when they are fabricated in a crossbar array structure. 8–16 Based on the desirable characteristics of the memristor crossbar array, various studies have demonstrated that the energy efficiency, hardware size, and inference speed can be effectively improved compared with the conventional CMOS-based processors 12,17,18…”

Section: Introductionmentioning

confidence: 99%