CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning
Abstract Extending Moore’s law by augmenting complementary-metal-oxide semiconductor (CMOS) transistors with emerging nanotechnologies (X) has become increasingly important. One important class of problems involve sampling-based Monte Carlo algorithms used in probabilistic machine learning, optimiza...
Main Authors: | , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Nature Portfolio
2024-03-01
|
Series: | Nature Communications |
Online Access: | https://doi.org/10.1038/s41467-024-46645-6 |
_version_ | 1797233439294881792 |
---|---|
author | Nihal Sanjay Singh Keito Kobayashi Qixuan Cao Kemal Selcuk Tianrui Hu Shaila Niazi Navid Anjum Aadit Shun Kanai Hideo Ohno Shunsuke Fukami Kerem Y. Camsari |
author_facet | Nihal Sanjay Singh Keito Kobayashi Qixuan Cao Kemal Selcuk Tianrui Hu Shaila Niazi Navid Anjum Aadit Shun Kanai Hideo Ohno Shunsuke Fukami Kerem Y. Camsari |
author_sort | Nihal Sanjay Singh |
collection | DOAJ |
description | Abstract Extending Moore’s law by augmenting complementary-metal-oxide semiconductor (CMOS) transistors with emerging nanotechnologies (X) has become increasingly important. One important class of problems involve sampling-based Monte Carlo algorithms used in probabilistic machine learning, optimization, and quantum simulation. Here, we combine stochastic magnetic tunnel junction (sMTJ)-based probabilistic bits (p-bits) with Field Programmable Gate Arrays (FPGA) to create an energy-efficient CMOS + X (X = sMTJ) prototype. This setup shows how asynchronously driven CMOS circuits controlled by sMTJs can perform probabilistic inference and learning by leveraging the algorithmic update-order-invariance of Gibbs sampling. We show how the stochasticity of sMTJs can augment low-quality random number generators (RNG). Detailed transistor-level comparisons reveal that sMTJ-based p-bits can replace up to 10,000 CMOS transistors while dissipating two orders of magnitude less energy. Integrated versions of our approach can advance probabilistic computing involving deep Boltzmann machines and other energy-based learning algorithms with extremely high throughput and energy efficiency. |
first_indexed | 2024-04-24T16:16:11Z |
format | Article |
id | doaj.art-141ef4996115451784de6c9033a24c3f |
institution | Directory Open Access Journal |
issn | 2041-1723 |
language | English |
last_indexed | 2024-04-24T16:16:11Z |
publishDate | 2024-03-01 |
publisher | Nature Portfolio |
record_format | Article |
series | Nature Communications |
spelling | doaj.art-141ef4996115451784de6c9033a24c3f2024-03-31T11:26:29ZengNature PortfolioNature Communications2041-17232024-03-011511910.1038/s41467-024-46645-6CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learningNihal Sanjay Singh0Keito Kobayashi1Qixuan Cao2Kemal Selcuk3Tianrui Hu4Shaila Niazi5Navid Anjum Aadit6Shun Kanai7Hideo Ohno8Shunsuke Fukami9Kerem Y. Camsari10Department of Electrical and Computer Engineering, University of California Santa BarbaraDepartment of Electrical and Computer Engineering, University of California Santa BarbaraDepartment of Electrical and Computer Engineering, University of California Santa BarbaraDepartment of Electrical and Computer Engineering, University of California Santa BarbaraDepartment of Electrical and Computer Engineering, University of California Santa BarbaraDepartment of Electrical and Computer Engineering, University of California Santa BarbaraDepartment of Electrical and Computer Engineering, University of California Santa BarbaraResearch Institute of Electrical Communication, Tohoku UniversityResearch Institute of Electrical Communication, Tohoku UniversityResearch Institute of Electrical Communication, Tohoku UniversityDepartment of Electrical and Computer Engineering, University of California Santa BarbaraAbstract Extending Moore’s law by augmenting complementary-metal-oxide semiconductor (CMOS) transistors with emerging nanotechnologies (X) has become increasingly important. One important class of problems involve sampling-based Monte Carlo algorithms used in probabilistic machine learning, optimization, and quantum simulation. Here, we combine stochastic magnetic tunnel junction (sMTJ)-based probabilistic bits (p-bits) with Field Programmable Gate Arrays (FPGA) to create an energy-efficient CMOS + X (X = sMTJ) prototype. This setup shows how asynchronously driven CMOS circuits controlled by sMTJs can perform probabilistic inference and learning by leveraging the algorithmic update-order-invariance of Gibbs sampling. We show how the stochasticity of sMTJs can augment low-quality random number generators (RNG). Detailed transistor-level comparisons reveal that sMTJ-based p-bits can replace up to 10,000 CMOS transistors while dissipating two orders of magnitude less energy. Integrated versions of our approach can advance probabilistic computing involving deep Boltzmann machines and other energy-based learning algorithms with extremely high throughput and energy efficiency.https://doi.org/10.1038/s41467-024-46645-6 |
spellingShingle | Nihal Sanjay Singh Keito Kobayashi Qixuan Cao Kemal Selcuk Tianrui Hu Shaila Niazi Navid Anjum Aadit Shun Kanai Hideo Ohno Shunsuke Fukami Kerem Y. Camsari CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning Nature Communications |
title | CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning |
title_full | CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning |
title_fullStr | CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning |
title_full_unstemmed | CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning |
title_short | CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning |
title_sort | cmos plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning |
url | https://doi.org/10.1038/s41467-024-46645-6 |
work_keys_str_mv | AT nihalsanjaysingh cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT keitokobayashi cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT qixuancao cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT kemalselcuk cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT tianruihu cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT shailaniazi cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT navidanjumaadit cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT shunkanai cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT hideoohno cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT shunsukefukami cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning AT keremycamsari cmosplusstochasticnanomagnetsenablingheterogeneouscomputersforprobabilisticinferenceandlearning |