Matches in SemOpenAlex for { <https://semopenalex.org/work/W4316022671> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W4316022671 abstract "Abstract Safety-critical sensory processing applications, like medical diagnosis, require making accurate decisions based on a small amount of noisy input data. For these applications, using Bayesian neural networks, able to quantify the uncertainty of the predictions, is a superior approach to using conventional artificial neural networks. However, because of the probabilistic nature of Bayesian neural networks, they can be computationally intensive to use for inference stage and thus not well suited for extreme-edge applications. An emerging idea to solve this problem is to use the intrinsic probabilistic nature of memristors to efficiently implement Bayesian neural network inference: the variability in the resistance of memristors would represent the probability distribution of weights in Bayesian neural networks. However, when using memristors, statistical effects follow the laws of device physics, whereas in Bayesian neural networks, those effects can take on arbitrary shapes. In this work, we overcome this difficulty by adopting a dedicated synapse architecture based on two memristors, and by training Bayesian neural networks with a dedicated variational inference technique that includes a “technological loss” to take into account specificities of memristor physics. This technique allowed us to program a two-layer Bayesian neural network on 75 physical crossbar arrays of 1,024 memristors, incorporating CMOS periphery circuitry to do in-memory computing, to classify arrhythmia in electrocardiograms. Our experimental neural network classified heartbeats with high accuracy, and estimated the certainty of its predictions. In the case of uncertain predictions, it differentiated between ambivalent heartbeats (aleatoric uncertainty), and heartbeats with never-seen patterns (epistemic uncertainty). We show that our technique can also be used with phase change memories, by employing a different “technological loss” term. The great advantage of this approach is its low energy consumption: we estimate an 800 times improvement in energy efficiency compared to a GPU performing the same task." @default.
- W4316022671 created "2023-01-14" @default.
- W4316022671 creator A5006951514 @default.
- W4316022671 creator A5020588660 @default.
- W4316022671 creator A5032162752 @default.
- W4316022671 creator A5039853755 @default.
- W4316022671 creator A5039888615 @default.
- W4316022671 creator A5046458712 @default.
- W4316022671 creator A5046559880 @default.
- W4316022671 creator A5047176060 @default.
- W4316022671 creator A5047747771 @default.
- W4316022671 creator A5049000221 @default.
- W4316022671 creator A5061839553 @default.
- W4316022671 creator A5063819347 @default.
- W4316022671 creator A5087577248 @default.
- W4316022671 date "2023-01-13" @default.
- W4316022671 modified "2023-10-18" @default.
- W4316022671 title "Bringing uncertainty quantification to the extreme-edge with memristor-based Bayesian neural networks" @default.
- W4316022671 doi "https://doi.org/10.21203/rs.3.rs-2458251/v1" @default.
- W4316022671 hasPublicationYear "2023" @default.
- W4316022671 type Work @default.
- W4316022671 citedByCount "0" @default.
- W4316022671 crossrefType "posted-content" @default.
- W4316022671 hasAuthorship W4316022671A5006951514 @default.
- W4316022671 hasAuthorship W4316022671A5020588660 @default.
- W4316022671 hasAuthorship W4316022671A5032162752 @default.
- W4316022671 hasAuthorship W4316022671A5039853755 @default.
- W4316022671 hasAuthorship W4316022671A5039888615 @default.
- W4316022671 hasAuthorship W4316022671A5046458712 @default.
- W4316022671 hasAuthorship W4316022671A5046559880 @default.
- W4316022671 hasAuthorship W4316022671A5047176060 @default.
- W4316022671 hasAuthorship W4316022671A5047747771 @default.
- W4316022671 hasAuthorship W4316022671A5049000221 @default.
- W4316022671 hasAuthorship W4316022671A5061839553 @default.
- W4316022671 hasAuthorship W4316022671A5063819347 @default.
- W4316022671 hasAuthorship W4316022671A5087577248 @default.
- W4316022671 hasBestOaLocation W43160226711 @default.
- W4316022671 hasConcept C107673813 @default.
- W4316022671 hasConcept C119857082 @default.
- W4316022671 hasConcept C127413603 @default.
- W4316022671 hasConcept C150072547 @default.
- W4316022671 hasConcept C154945302 @default.
- W4316022671 hasConcept C160234255 @default.
- W4316022671 hasConcept C24326235 @default.
- W4316022671 hasConcept C2776214188 @default.
- W4316022671 hasConcept C33724603 @default.
- W4316022671 hasConcept C41008148 @default.
- W4316022671 hasConcept C49937458 @default.
- W4316022671 hasConcept C50644808 @default.
- W4316022671 hasConceptScore W4316022671C107673813 @default.
- W4316022671 hasConceptScore W4316022671C119857082 @default.
- W4316022671 hasConceptScore W4316022671C127413603 @default.
- W4316022671 hasConceptScore W4316022671C150072547 @default.
- W4316022671 hasConceptScore W4316022671C154945302 @default.
- W4316022671 hasConceptScore W4316022671C160234255 @default.
- W4316022671 hasConceptScore W4316022671C24326235 @default.
- W4316022671 hasConceptScore W4316022671C2776214188 @default.
- W4316022671 hasConceptScore W4316022671C33724603 @default.
- W4316022671 hasConceptScore W4316022671C41008148 @default.
- W4316022671 hasConceptScore W4316022671C49937458 @default.
- W4316022671 hasConceptScore W4316022671C50644808 @default.
- W4316022671 hasLocation W43160226711 @default.
- W4316022671 hasOpenAccess W4316022671 @default.
- W4316022671 hasPrimaryLocation W43160226711 @default.
- W4316022671 hasRelatedWork W1563448620 @default.
- W4316022671 hasRelatedWork W2072389162 @default.
- W4316022671 hasRelatedWork W2210219486 @default.
- W4316022671 hasRelatedWork W2532368719 @default.
- W4316022671 hasRelatedWork W2561451072 @default.
- W4316022671 hasRelatedWork W2574982804 @default.
- W4316022671 hasRelatedWork W2783359940 @default.
- W4316022671 hasRelatedWork W2963058055 @default.
- W4316022671 hasRelatedWork W3207575386 @default.
- W4316022671 hasRelatedWork W4286903091 @default.
- W4316022671 isParatext "false" @default.
- W4316022671 isRetracted "false" @default.
- W4316022671 workType "article" @default.