【摘 要】
:
The hardware implementation of neural networks based on memristor crossbar array provides a promising paradigm for neuromorphic computing.However,the existence of memristor conductance drift harms the reliability of the deployed neural network,which serio
【机 构】
:
College of Electronics and Information Engineering,Southwest University,Chongqing 400715,China;Colle
论文部分内容阅读
The hardware implementation of neural networks based on memristor crossbar array provides a promising paradigm for neuromorphic computing.However,the existence of memristor conductance drift harms the reliability of the deployed neural network,which seriously hinders the practical application of memristor-based neuromorphic computing.In this paper,the impact of different types of conductance drift on the weight realized by memristors is investigated and analyzed.Then,utilizing the weight uncertainty introduced by conductance drift,we propose a weight optimization method based on the Bayesian neural network,which can greatly improve the network performance.Furthermore,an ensemble approach is pro-posed to enhance network reliability without increasing training cost or crossbar array resources.Finally,the effectiveness of the proposed scheme is verified through a series of experiments.In addition,the proposed scheme can be easily integrated into the implementation of neuromorphic computing,which can provide a better guarantee for its large-scale application.
其他文献
In traditional von Neumann computing architectures,the essential transfer of data between the processor and memory hierarchies limits the computational efficiency of next-generation system-on-a-chip.The emerging in-memory computing (IMC) approach addresse
Dear editor,rnAs an important mathematical model for the elucidation,analysis,and control of gene regulatory networks (GRNs),logical networks have attracted considerable attention from scientists in numerous fields of study.From a mathematical point of vi
Memristor based computing-in-memory chips have shown the potentials to accelerate deep neural networks with high energy efficiency.Due to the inherent filament-based conductive mechanism of the memristor,the reading and writing noises are hard to eliminat
Dear editor,rnThe fictitious play (FP) was initially proposed by Brown[1]in 1951 as a handy learning rule.The main idea of FP can be described as follows:In a game process,the player who adopts FP assumes that the others choose their strate-gies randomly
Protected satellite communications (SatComs) exhibit specific characteristics such as security,intelligence,anti-jamming,and nuclear disaster survivability.They constitute one of the key research top-ics in modern military communications and have become t
Nowadays,massive open online courses (MOOCs),are at-tracting widespread interest as an alternative education model.Several MOOCs platforms,such as Coursera,edX,and Udacity have been built and they provide low-cost op-portunities for anyone who needs to ac
Dear editor,rnMany emerging non-volatile memories (NVM),such as re-sistive random access memory (RRAM)[1],phase-change memory (PCM)[2],and ferroelectric RAM (FeRAM)[3],together with the conventional flash memory[4,5],have demonstrated their good capabilit
Dear editor,rnSAT solvers,based on heuristic algorithms,are used to solve Boolean satisfiability (SAT) problems.Satisfiability mod-ulo theories (SMT) problem is a decision problem concerned with the satisfiability of a logical formula;it is expressed as a
Detection of interictal epileptic discharges (IED) events in the EEG recordings is a critical indicator for detecting and diagnosing epileptic seizures.We propose a key technology to extract the most important features related to epileptic seizures and id
Dear editor,rnRecently,considerable attention has been devoted to out-put feedback control and practical tracking of nonlinear sys-tems[1,2].Unfortunately,limitations of sensor techniques can cause sensitivity errors in practical environments.For example,