Matches in SemOpenAlex for { <https://semopenalex.org/work/W4310525971> ?p ?o ?g. }
Showing items 1 to 63 of
63
with 100 items per page.
- W4310525971 abstract "The hardware implementation of neural networksbased on memristor crossbar array provides a promisingparadigm for neuromorphic computing. However, the existenceof memristor conductance drift harms the reliability of thedeployed neural network, which seriously hinders the practicalapplication of memristor-based neuromorphic computing. In thispaper, the impact of different types of conductance drift onthe weight realized by memristors is investigated and analyzed.Then, utilizing the weight uncertainty introduced by conductancedrift, we propose a weight optimization method based on theBayesian neural network, which can greatly improve the networkperformance. Furthermore, an ensemble approach is proposedto enhance network reliability without increasing training costor crossbar array resources. Finally, the effectiveness of theproposed scheme is verified through a series of experiments. Inaddition, the proposed scheme can be easily integrated into theimplementation of neuromorphic computing, which can providea better guarantee for its large-scale application. For preventingmemristor defects from degrading edge intelligence performance,chip-in-the-loop training can be useful when training memristorcrossbars. Another undesirable effect in memristor crossbars isparasitic resistances such as source, line, and neuron resistance,which worsens as crossbar size increases. Various circuit andsoftware techniques can compensate for parasitic resistances likesource, line, and neuron resistance. Finally, we discuss an energy-efficient programming method for updating synaptic weights inmemristor crossbars, which is needed for learning the edgedevices." @default.
- W4310525971 created "2022-12-11" @default.
- W4310525971 creator A5089329858 @default.
- W4310525971 date "2022-11-30" @default.
- W4310525971 modified "2023-09-24" @default.
- W4310525971 title "Quantization and parasitic resistance correction for enhancing reliability against conductance drift with Memristor neural networks" @default.
- W4310525971 doi "https://doi.org/10.31237/osf.io/hqpn9" @default.
- W4310525971 hasPublicationYear "2022" @default.
- W4310525971 type Work @default.
- W4310525971 citedByCount "0" @default.
- W4310525971 crossrefType "posted-content" @default.
- W4310525971 hasAuthorship W4310525971A5089329858 @default.
- W4310525971 hasBestOaLocation W43105259712 @default.
- W4310525971 hasConcept C11413529 @default.
- W4310525971 hasConcept C121332964 @default.
- W4310525971 hasConcept C127413603 @default.
- W4310525971 hasConcept C150072547 @default.
- W4310525971 hasConcept C151927369 @default.
- W4310525971 hasConcept C154945302 @default.
- W4310525971 hasConcept C163258240 @default.
- W4310525971 hasConcept C24326235 @default.
- W4310525971 hasConcept C28855332 @default.
- W4310525971 hasConcept C29984679 @default.
- W4310525971 hasConcept C41008148 @default.
- W4310525971 hasConcept C43214815 @default.
- W4310525971 hasConcept C50644808 @default.
- W4310525971 hasConcept C62520636 @default.
- W4310525971 hasConcept C66949984 @default.
- W4310525971 hasConcept C76155785 @default.
- W4310525971 hasConceptScore W4310525971C11413529 @default.
- W4310525971 hasConceptScore W4310525971C121332964 @default.
- W4310525971 hasConceptScore W4310525971C127413603 @default.
- W4310525971 hasConceptScore W4310525971C150072547 @default.
- W4310525971 hasConceptScore W4310525971C151927369 @default.
- W4310525971 hasConceptScore W4310525971C154945302 @default.
- W4310525971 hasConceptScore W4310525971C163258240 @default.
- W4310525971 hasConceptScore W4310525971C24326235 @default.
- W4310525971 hasConceptScore W4310525971C28855332 @default.
- W4310525971 hasConceptScore W4310525971C29984679 @default.
- W4310525971 hasConceptScore W4310525971C41008148 @default.
- W4310525971 hasConceptScore W4310525971C43214815 @default.
- W4310525971 hasConceptScore W4310525971C50644808 @default.
- W4310525971 hasConceptScore W4310525971C62520636 @default.
- W4310525971 hasConceptScore W4310525971C66949984 @default.
- W4310525971 hasConceptScore W4310525971C76155785 @default.
- W4310525971 hasLocation W43105259711 @default.
- W4310525971 hasLocation W43105259712 @default.
- W4310525971 hasLocation W43105259713 @default.
- W4310525971 hasOpenAccess W4310525971 @default.
- W4310525971 hasPrimaryLocation W43105259711 @default.
- W4310525971 hasRelatedWork W2155954834 @default.
- W4310525971 hasRelatedWork W2588565308 @default.
- W4310525971 hasRelatedWork W2800431044 @default.
- W4310525971 hasRelatedWork W2900722538 @default.
- W4310525971 hasRelatedWork W2954452552 @default.
- W4310525971 hasRelatedWork W2997454382 @default.
- W4310525971 hasRelatedWork W3157876397 @default.
- W4310525971 hasRelatedWork W4206933960 @default.
- W4310525971 hasRelatedWork W4293733579 @default.
- W4310525971 hasRelatedWork W4310525971 @default.
- W4310525971 isParatext "false" @default.
- W4310525971 isRetracted "false" @default.
- W4310525971 workType "article" @default.