Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

RESISTIVE RANDOM-ACCESS MEMORY

One-shot learning goes 3D

Nature Electronics volume 4, pages 866–867 (2021)Cite this article

Subjects

A monolithic three-dimensional integrated system based on CMOS logic, compute-in-memory and associative memory can be used to efficiently implement one-shot learning.

Deep neural networks are highly effective at processing visual and speech data. However, they typically require an immense amount of training data and processing power to achieve state-of-the-art accuracy1. They also struggle to adapt to new classes of sample data and must be re-trained to extract information2. One promising approach to alleviate the training data requirements and enable adaptability is to use memory augmented neural networks3, where an attentional memory is added to store the features extracted from the neural network. An important application of such systems is one-shot (or few-shot) learning4, which requires only one (or a few) training sample(s) for each class. Here the stored features in the attentional memory are retrieved for predictions based on their distances (for example, cosine distance and Hamming distance) to the new query sample. Hardware implementation of one-/few-shot learning using conventional von Neumann architectures is challenging due to energy-consuming and long-latency data movement between the computing unit and memory5. It is, instead, preferable to perform the computation required by the deep neural networks and attentional memory directly within memory6. Reporting at the 2021 IEEE International Electron Devices Meeting in San Francisco, Jianshi Tang, Huaqiang Wu and colleagues now demonstrate a monolithic three-dimensional (3D) integrated system that is based on resistive random-access memory (RRAM) and can compactly and energy efficiently implement one-/few-shot learning7.

This is a preview of subscription content, access via your institution

Access options

Rent or buy this article

Get just this article for as long as you need it

$39.95

Learn more

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: A monolithic 3D integrated one-shot learning system.

References

  1. Lecun, Y., Bengio, Y. & Hinton, G. Nature 521, 436–444 (2015).

    Article  Google Scholar 

  2. Kirkpatrick, J. et al. Proc. Natl Acad. Sci. USA 114, 3521–3526 (2017).

    Article  MathSciNet  Google Scholar 

  3. Graves, A., Wayne, G. & Danihelka, I. Preprint at https://arxiv.org/abs/1410.5401 (2014).

  4. Vinyals, O., Blundell, C., Lillicrap, T., Kavukcuoglu, K. & Wierstra, D. In Advances in Neural Information Processing Systems 29, 3637–3645 (2016).

  5. Ielmini, D. & Wong, H. S. P. Nat. Electron. 1, 333–343 (2018).

    Article  Google Scholar 

  6. Ni, K. et al. Nat. Electron. 2, 521–529 (2019).

    Article  Google Scholar 

  7. Li, Y. et al. In Proc. 2021 IEEE Int. Electron Devices Meeting (in the press); https://ieee-iedm.org/program/

  8. Batude, P. et al. In Proc. 2017 IEEE Int. Electron Devices Meeting 3.1.1–3.1.4 (IEEE, 2017); https://doi.org/10.1109/IEDM.2017.8268316

  9. Yin, X. et al. IEEE Trans. Electron Devices 67, 2785–2792 (2020).

    Article  Google Scholar 

  10. Kazemi, A. et al. In Proc. 2021 Design, Automation Test in Europe Conference Exhibition (DATE) 1084–1089 (IEEE, 2021); https://doi.org/10.23919/DATE51398.2021.9474025

Download references

Author information

Authors and Affiliations

  1. Rochester Institute of Technology, Rochester, NY, USA

    Zijian Zhao, Shan Deng, Zhouhang Jiang & Kai Ni

Authors
  1. Zijian Zhao
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Shan Deng
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Zhouhang Jiang
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Kai Ni
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Kai Ni.

Ethics declarations

Competing interests

The authors declare no competing interests.

Rights and permissions

Reprints and Permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhao, Z., Deng, S., Jiang, Z. et al. One-shot learning goes 3D. Nat Electron 4, 866–867 (2021). https://doi.org/10.1038/s41928-021-00692-9

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41928-021-00692-9

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing