|
1
|
Singh NS, Kobayashi K, Cao Q, Selcuk K, Hu T, Niazi S, Aadit NA, Kanai S, Ohno H, Fukami S, Camsari KY. CMOS plus stochastic nanomagnets enabling heterogeneous computers for probabilistic inference and learning. Nat Commun 2024; 15:2685. [PMID: 38538599 PMCID: PMC10973401 DOI: 10.1038/s41467-024-46645-6] [Citation(s) in RCA: 5] [Impact Index Per Article: 5.0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/03/2023] [Accepted: 03/05/2024] [Indexed: 11/12/2024] Open
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.
Collapse
Affiliation(s)
- Nihal Sanjay Singh
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
| | - Keito Kobayashi
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
- Research Institute of Electrical Communication, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- Graduate School of Engineering, Tohoku University, 6-6 Aramaki Aza Aoba, Aoba-ku, Sendai, 980-0845, Japan
| | - Qixuan Cao
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
| | - Kemal Selcuk
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
| | - Tianrui Hu
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
| | - Shaila Niazi
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
| | - Navid Anjum Aadit
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA
| | - Shun Kanai
- Research Institute of Electrical Communication, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- Graduate School of Engineering, Tohoku University, 6-6 Aramaki Aza Aoba, Aoba-ku, Sendai, 980-0845, Japan
- WPI Advanced Institute for Materials Research (WPI-AIMR), Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- Center for Science and Innovation in Spintronics (CSIS), Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- PRESTO, Japan Science and Technology Agency (JST), Kawaguchi, 332-0012, Japan
- Division for the Establishment of Frontier Sciences of Organization for Advanced Studies at Tohoku University, Tohoku University, Sendai, 980-8577, Japan
- National Institutes for Quantum Science and Technology, Takasaki, 370-1207, Japan
| | - Hideo Ohno
- Research Institute of Electrical Communication, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- WPI Advanced Institute for Materials Research (WPI-AIMR), Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- Center for Science and Innovation in Spintronics (CSIS), Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan
- Center for Innovative Integrated Electronic Systems (CIES), Tohoku University, 468-1 Aramaki Aza Aoba, Aoba-ku, Sendai, 980-0845, Japan
| | - Shunsuke Fukami
- Research Institute of Electrical Communication, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan.
- Graduate School of Engineering, Tohoku University, 6-6 Aramaki Aza Aoba, Aoba-ku, Sendai, 980-0845, Japan.
- WPI Advanced Institute for Materials Research (WPI-AIMR), Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan.
- Center for Science and Innovation in Spintronics (CSIS), Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai, 980-8577, Japan.
- Center for Innovative Integrated Electronic Systems (CIES), Tohoku University, 468-1 Aramaki Aza Aoba, Aoba-ku, Sendai, 980-0845, Japan.
- Inamori Research Institute of Science (InaRIS), Kyoto, 600-8411, Japan.
| | - Kerem Y Camsari
- Department of Electrical and Computer Engineering, University of California Santa Barbara, Santa Barbara, 93106, CA, USA.
| |
Collapse
|
|
2
|
Local bifurcation with spin-transfer torque in superparamagnetic tunnel junctions. Nat Commun 2022; 13:4079. [PMID: 35835780 PMCID: PMC9283488 DOI: 10.1038/s41467-022-31788-1] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.3] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/20/2021] [Accepted: 07/04/2022] [Indexed: 11/18/2022] Open
Abstract
Modulation of the energy landscape by external perturbations governs various thermally-activated phenomena, described by the Arrhenius law. Thermal fluctuation of nanoscale magnetic tunnel junctions with spin-transfer torque (STT) shows promise for unconventional computing, whereas its rigorous representation, based on the Néel-Arrhenius law, has been controversial. In particular, the exponents for thermally-activated switching rate therein, have been inaccessible with conventional thermally-stable nanomagnets with decade-long retention time. Here we approach the Néel-Arrhenius law with STT utilising superparamagnetic tunnel junctions that have high sensitivity to external perturbations and determine the exponents through several independent measurements including homodyne-detected ferromagnetic resonance, nanosecond STT switching, and random telegraph noise. Furthermore, we show that the results are comprehensively described by a concept of local bifurcation observed in various physical systems. The findings demonstrate the capability of superparamagnetic tunnel junction as a useful tester for statistical physics as well as sophisticated engineering of probabilistic computing hardware with a rigorous mathematical foundation. There has been much interest in using the probabilistic switching of magnetic tunnel junctions in unconventional computing, but to do so requires a detailed understanding of this switching. Here, Funatsu et al rigorously determine the switching exponents in superparamagnetic tunnel junctions.
Collapse
|