File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Uncertainty quantification via a memristor Bayesian deep neural network for risk-sensitive reinforcement learning

TitleUncertainty quantification via a memristor Bayesian deep neural network for risk-sensitive reinforcement learning
Authors
Issue Date2023
Citation
Nature Machine Intelligence, 2023, v. 5, n. 7, p. 714-723 How to Cite?
AbstractMany advanced artificial intelligence tasks, such as policy optimization, decision making and autonomous navigation, demand high-bandwidth data transfer and probabilistic computing, posing great challenges for conventional computing hardware. Since digital computers based on the von Neumann architecture are good at precise and deterministic computing, their computing efficiency is limited by the high cost of both data transfer between memory and computing units and massive random number generation. Here we develop a stochastic computation-in-memory computing system that can efficiently perform both in situ random number generation and computation based on the nanoscale physical behaviour of memristors. This system is constructed based on a hardware-implemented multiple-memristor-array system. To demonstrate its functionality and efficiency, we implement a typical risk-sensitive reinforcement learning task, namely the storm coast task, with a four-layer Bayesian deep neural network. The computing system efficiently decomposes aleatoric and epistemic uncertainties by exploiting the inherent stochasticity of memristor. Compared with the conventional digital computer, our memristor-based system achieves a 10 times higher speed and 150 times higher energy efficiency in uncertainty decomposition. This stochastic computation-in-memory computing system paves the way for high-speed and energy-efficient implementation of various probabilistic artificial intelligence algorithms.
Persistent Identifierhttp://hdl.handle.net/10722/334960
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorLin, Yudeng-
dc.contributor.authorZhang, Qingtian-
dc.contributor.authorGao, Bin-
dc.contributor.authorTang, Jianshi-
dc.contributor.authorYao, Peng-
dc.contributor.authorLi, Chongxuan-
dc.contributor.authorHuang, Shiyu-
dc.contributor.authorLiu, Zhengwu-
dc.contributor.authorZhou, Ying-
dc.contributor.authorLiu, Yuyi-
dc.contributor.authorZhang, Wenqiang-
dc.contributor.authorZhu, Jun-
dc.contributor.authorQian, He-
dc.contributor.authorWu, Huaqiang-
dc.date.accessioned2023-10-20T06:52:00Z-
dc.date.available2023-10-20T06:52:00Z-
dc.date.issued2023-
dc.identifier.citationNature Machine Intelligence, 2023, v. 5, n. 7, p. 714-723-
dc.identifier.urihttp://hdl.handle.net/10722/334960-
dc.description.abstractMany advanced artificial intelligence tasks, such as policy optimization, decision making and autonomous navigation, demand high-bandwidth data transfer and probabilistic computing, posing great challenges for conventional computing hardware. Since digital computers based on the von Neumann architecture are good at precise and deterministic computing, their computing efficiency is limited by the high cost of both data transfer between memory and computing units and massive random number generation. Here we develop a stochastic computation-in-memory computing system that can efficiently perform both in situ random number generation and computation based on the nanoscale physical behaviour of memristors. This system is constructed based on a hardware-implemented multiple-memristor-array system. To demonstrate its functionality and efficiency, we implement a typical risk-sensitive reinforcement learning task, namely the storm coast task, with a four-layer Bayesian deep neural network. The computing system efficiently decomposes aleatoric and epistemic uncertainties by exploiting the inherent stochasticity of memristor. Compared with the conventional digital computer, our memristor-based system achieves a 10 times higher speed and 150 times higher energy efficiency in uncertainty decomposition. This stochastic computation-in-memory computing system paves the way for high-speed and energy-efficient implementation of various probabilistic artificial intelligence algorithms.-
dc.languageeng-
dc.relation.ispartofNature Machine Intelligence-
dc.titleUncertainty quantification via a memristor Bayesian deep neural network for risk-sensitive reinforcement learning-
dc.typeArticle-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1038/s42256-023-00680-y-
dc.identifier.scopuseid_2-s2.0-85162980659-
dc.identifier.volume5-
dc.identifier.issue7-
dc.identifier.spage714-
dc.identifier.epage723-
dc.identifier.eissn2522-5839-
dc.identifier.isiWOS:001014470500001-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats