---
title: "Matching Networks for One Shot Learning — Vinyals 2016"
description: "The authors from Google DeepMind propose Matching Networks. 30 references, cited by 343 papers here."
authors: ["Oriol Vinyals", "Charles Blundell", "Timothy Lillicrap", "Koray Kavukcuoglu", "Daan Wierstra"]
year: 2016
venue: "arXiv:1606.04080"
source: "https://arxiv.org/abs/1606.04080"
doi: "10.48550/arxiv.1606.04080"
arxiv: "1606.04080"
licence: "arXiv — non-exclusive licence to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/vinyals-2016-matching-networks-one-shot-learning/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/vinyals-2016-matching-networks-one-shot-learning/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2016](https://md.research.elamur.ai/free/en/papers/2016/) / Matching Networks for One Shot Learning

# Matching Networks for One Shot Learning

**Authors:** [Oriol Vinyals](https://md.research.elamur.ai/free/en/authors/oriol-vinyals/), Charles Blundell, [Timothy Lillicrap](https://md.research.elamur.ai/free/en/authors/timothy-lillicrap/), [Koray Kavukcuoglu](https://md.research.elamur.ai/free/en/authors/koray-kavukcuoglu/) and [Daan Wierstra](https://md.research.elamur.ai/free/en/authors/daan-wierstra/) · **Year:** 2016 · **Venue:** arXiv:1606.04080 · **Licence:** [arXiv — non-exclusive licence to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/1606.04080) · [PDF](https://arxiv.org/pdf/1606.04080) · [DOI 10.48550/arxiv.1606.04080](https://doi.org/10.48550/arxiv.1606.04080) · [arXiv 1606.04080](https://arxiv.org/abs/1606.04080) · [OpenAlex](https://openalex.org/W2432717477)

**Citations:** 854 citations (OpenAlex, 2026-10-14) · [343](https://md.research.elamur.ai/free/en/papers/vinyals-2016-matching-networks-one-shot-learning/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary

The paper addresses one-shot learning: how to train a neural network to recognize a new class when only one or a few labeled examples are available. Classical deep learning requires large datasets and lengthy fine-tuning for this. The authors from Google DeepMind propose Matching Networks. This model takes a small labeled support set and a new example and immediately outputs probabilities over the labels in that set. The prediction is a weighted sum of the labels of the support examples. The weights are given by an attention mechanism: a softmax over the cosine similarity of embeddings. The embeddings can take the whole set into account thanks to an LSTM. The second idea is episodic training. During training, the model receives a new small task with several classes each time, so training mimics test conditions. As a result, no fine-tuning is needed for new classes. On Omniglot and ImageNet, the method improved one-shot classification accuracy by roughly 5–6 percentage points over competing approaches. The authors also proposed new benchmarks, including a language task on Penn Treebank. This work became one of the foundational approaches in meta-learning.

## Key points

- **Results on images:** on ImageNet, accuracy rose from 87.6% to 93.2%, and on Omniglot from 88.0% to 93.8%, compared with the best alternative methods.
- **Architecture:** Full Context Embeddings. The elements of the support set are encoded by a bidirectional LSTM, and the test example is processed by an LSTM with attention over the whole set for a fixed number of steps K. The model combines ideas from kNN, kernel density estimation, and networks with external memory (memory networks, pointer networks).
- **Language task on Penn Treebank:** a word is missing from a query sentence, and it must be chosen from 5 candidates, given example sentences with the same gaps. The words are split into 9000 training and 1000 test words. Accuracy is 32.4%, 36.1%, and 38.2% for k = 1, 2, 3. Chance level is 20%, while an oracle LSTM-LM that has seen all the data reaches 72.8%.
- **New benchmarks:** one-shot tasks on the full ImageNet, on a reduced version of ImageNet for fast experiments, and a language task on PTB.
- **Limitations:** the cost of each gradient step grows with the size of the support set. The model is also weaker on fine-grained classes, such as the dog breeds subtask on ImageNet.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Google](https://md.research.elamur.ai/free/en/companies/google/) — Company, Mountain View, United States · 430 works · 424,363 citations
- [Google DeepMind](https://md.research.elamur.ai/free/en/companies/google-deepmind/) — Company, London, United Kingdom · 56 works · 78,880 citations

📚 Shelves (3): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works

💡 Concepts (2): [💡 Natural language processing](https://md.research.elamur.ai/free/en/concepts/natural-language-processing/) · 131 works · [💡 Few-shot learning](https://md.research.elamur.ai/free/en/concepts/few-shot-learning/) · 45 works

## Cite

Cite as: Oriol Vinyals, Charles Blundell, Timothy Lillicrap, Koray Kavukcuoglu and Daan Wierstra. 2016. Matching Networks for One Shot Learning. arXiv:1606.04080.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{vinyals2016matching,
  title = {{Matching Networks for One Shot Learning}},
  author = {Oriol Vinyals, Charles Blundell, Timothy Lillicrap, Koray Kavukcuoglu and Daan Wierstra},
  year = {2016},
  doi = {10.48550/arxiv.1606.04080},
  eprint = {1606.04080},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1606.04080}
}
```

- [← previous Aspect Level Sentiment Classification with Deep Memory Network](https://md.research.elamur.ai/free/en/papers/tang-2016-aspect-level-sentiment-classification-deep/)
- [next → MS MARCO: A Human Generated MAchine Reading COmprehension Dataset](https://md.research.elamur.ai/free/en/papers/bajaj-2016-ms-marco-human-generated/)
- [↑ back to papers of 2016](https://md.research.elamur.ai/free/en/papers/2016/)

## References

30 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. C Atkeson, A Moore, and S Schaal. Locally weighted learning. Artificial Intelligence Review, 1997. [find](https://scholar.google.com/scholar?q=C+Atkeson%2C+A+Moore%2C+and+S+Schaal%2E+Locally+weighted+learning%2E+Artificial+Intelligence+Review%2C+1997%2E)
2. D Bahdanau, K Cho, and Y Bengio. [Neural machine translation by jointly learning to align and translate](https://md.research.elamur.ai/free/en/papers/bahdanau-2014-neural-machine-translation-jointly-learning/). ICLR, 2014.
3. J Donahue, Y Jia, O Vinyals, J Hoffman, N Zhang, E Tzeng, and T Darrell. Decaf: A deep convolutional activation feature for generic visual recognition. In ICML, 2014. [find](https://scholar.google.com/scholar?q=J+Donahue%2C+Y+Jia%2C+O+Vinyals%2C+J+Hoffman%2C+N+Zhang%2C+E+Tzeng%2C+and+T+Darrell%2E+Decaf%3A+A+deep+convolutional+activation+feature+for+generic+visual+recognition%2E+In+ICML%2C+2014%2E)
4. A Graves, G Wayne, and I Danihelka. Neural turing machines. arXiv preprint arXiv:1410.5401, 2014. [find](https://scholar.google.com/scholar?q=A+Graves%2C+G+Wayne%2C+and+I+Danihelka%2E+Neural+turing+machines%2E+arXiv+preprint+arXiv%3A1410%2E5401%2C+2014%2E)
5. K Hermann, T Kocisky, E Grefenstette, L Espeholt, W Kay, M Suleyman, and P Blunsom. [Teaching machines to read and comprehend](https://md.research.elamur.ai/free/en/papers/hermann-2015-teaching-machines-read-comprehend/). In NIPS, 2015.
6. F Hill, A Bordes, S Chopra, and J Weston. The goldilocks principle: Reading children’s books with explicit memory representations. arXiv preprint arXiv:1511.[02301, 2015](https://md.research.elamur.ai/free/en/papers/hill-2015-goldilocks-principle-reading-children-s/).
7. G Hinton et al. Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups. Signal Processing Magazine, IEEE, 2012. [find](https://scholar.google.com/scholar?q=G+Hinton+et+al%2E+Deep+neural+networks+for+acoustic+modeling+in+speech+recognition%3A+The+shared+views+of+four+research+groups%2E+Signal+Processing+Magazine%2C+IEEE%2C+2012%2E)
8. S Hochreiter and J Schmidhuber. Long short-term memory. Neural computation, 1997. [find](https://scholar.google.com/scholar?q=S+Hochreiter+and+J+Schmidhuber%2E+Long+short%2Dterm+memory%2E+Neural+computation%2C+1997%2E)
9. E Hoffer and N Ailon. Deep metric learning using triplet network. Similarity-Based Pattern Recognition, 2015. [find](https://scholar.google.com/scholar?q=E+Hoffer+and+N+Ailon%2E+Deep+metric+learning+using+triplet+network%2E+Similarity%2DBased+Pattern+Recognition%2C+2015%2E)
10. S Ioffe and C Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint arXiv:1502.03167, 2015. [find](https://scholar.google.com/scholar?q=S+Ioffe+and+C+Szegedy%2E+Batch+normalization%3A+Accelerating+deep+network+training+by+reducing+internal+covariate+shift%2E+arXiv+preprint+arXiv%3A1502%2E03167%2C+2015%2E)
11. G Koch, R Zemel, and R Salakhutdinov. Siamese neural networks for one-shot image recognition. In ICML Deep Learning workshop, 2015. [find](https://scholar.google.com/scholar?q=G+Koch%2C+R+Zemel%2C+and+R+Salakhutdinov%2E+Siamese+neural+networks+for+one%2Dshot+image+recognition%2E+In+ICML+Deep+Learning+workshop%2C+2015%2E)
12. A Krizhevsky and G Hinton. Convolutional deep belief networks on cifar-10. Unpublished, 2010. [find](https://scholar.google.com/scholar?q=A+Krizhevsky+and+G+Hinton%2E+Convolutional+deep+belief+networks+on+cifar%2D10%2E+Unpublished%2C+2010%2E)
13. A Krizhevsky, I Sutskever, and G Hinton. [Imagenet classification with deep convolutional neural networks](https://md.research.elamur.ai/free/en/papers/krizhevsky-2012-imagenet-classification-deep-convolutional/). In NIPS, 2012.
14. BM Lake, R Salakhutdinov, J Gross, and J Tenenbaum. One shot learning of simple visual concepts. In CogSci, 2011. [find](https://scholar.google.com/scholar?q=BM+Lake%2C+R+Salakhutdinov%2C+J+Gross%2C+and+J+Tenenbaum%2E+One+shot+learning+of+simple+visual+concepts%2E+In+CogSci%2C+2011%2E)
15. MP Marcus, MA Marcinkiewicz, and B Santorini. [Building a large annotated corpus of english: The penn treebank](https://md.research.elamur.ai/free/en/papers/marcus-1993-building-large-annotated-corpus-english/). Computational linguistics, 1993.
16. T Mikolov, M Karafiát, L Burget, J Cernockỳ, and S Khudanpur. Recurrent neural network based language model. In INTERSPEECH, 2010. [find](https://scholar.google.com/scholar?q=T+Mikolov%2C+M+Karafi%C3%A1t%2C+L+Burget%2C+J+Cernocky%CC%80%2C+and+S+Khudanpur%2E+Recurrent+neural+network+based+language+model%2E+In+INTERSPEECH%2C+2010%2E)
17. M Norouzi, T Mikolov, S Bengio, Y Singer, J Shlens, A Frome, G Corrado, and J Dean. Zero-shot learning by convex combination of semantic embeddings. arXiv preprint arXiv:1312.5650, 2013. — [arXiv:1312.5650](https://arxiv.org/abs/1312.5650) [find](https://scholar.google.com/scholar?q=M+Norouzi%2C+T+Mikolov%2C+S+Bengio%2C+Y+Singer%2C+J+Shlens%2C+A+Frome%2C+G+Corrado%2C+and+J+Dean%2E+Zero%2Dshot+learning+by+convex+combination+of+semantic+embeddings%2E+arXiv+preprint+arXiv%3A1312%2E5650%2C+2013%2E+%E2%80%94+%5BarXiv%3A1312)
18. S Roweis, G Hinton, and R Salakhutdinov. Neighbourhood component analysis. NIPS, 2004. [find](https://scholar.google.com/scholar?q=S+Roweis%2C+G+Hinton%2C+and+R+Salakhutdinov%2E+Neighbourhood+component+analysis%2E+NIPS%2C+2004%2E)
19. O Russakovsky, J Deng, H Su, J Krause, S Satheesh, S Ma, Z Huang, A Karpathy, A Khosla, M Bernstein, A Berg, and L Fei-Fei. [ImageNet Large Scale Visual Recognition Challenge](https://md.research.elamur.ai/free/en/papers/russakovsky-2014-imagenet-large-scale-visual-recognition/). IJCV, 2015.
20. R Salakhutdinov and G Hinton. Learning a nonlinear embedding by preserving class neighbourhood structure. In AISTATS, 2007. [find](https://scholar.google.com/scholar?q=R+Salakhutdinov+and+G+Hinton%2E+Learning+a+nonlinear+embedding+by+preserving+class+neighbourhood+structure%2E+In+AISTATS%2C+2007%2E)
21. A Santoro, S Bartunov, M Botvinick, D Wierstra, and T Lillicrap. Meta-learning with memory-augmented neural networks. In ICML, 2016. [find](https://scholar.google.com/scholar?q=A+Santoro%2C+S+Bartunov%2C+M+Botvinick%2C+D+Wierstra%2C+and+T+Lillicrap%2E+Meta%2Dlearning+with+memory%2Daugmented+neural+networks%2E+In+ICML%2C+2016%2E)
22. K Simonyan and A Zisserman. [Very deep convolutional networks for large-scale image recognition](https://md.research.elamur.ai/free/en/papers/simonyan-2014-very-deep-convolutional-networks-large/). arXiv preprint arXiv:1409.1556, 2014.
23. I Sutskever, O Vinyals, and QV Le. [Sequence to sequence learning with neural networks](https://md.research.elamur.ai/free/en/papers/sutskever-2014-sequence-sequence-learning-neural-networks/). In NIPS, 2014.
24. C Szegedy, W Liu, Y Jia, P Sermanet, S Reed, D Anguelov, D Erhan, V Vanhoucke, and A Rabinovich. Going deeper with convolutions. In CVPR, 2015. [find](https://scholar.google.com/scholar?q=C+Szegedy%2C+W+Liu%2C+Y+Jia%2C+P+Sermanet%2C+S+Reed%2C+D+Anguelov%2C+D+Erhan%2C+V+Vanhoucke%2C+and+A+Rabinovich%2E+Going+deeper+with+convolutions%2E+In+CVPR%2C+2015%2E)
25. C Szegedy, V Vanhoucke, S Ioffe, J Shlens, and Z Wojna. Rethinking the inception architecture for computer vision. arXiv preprint arXiv:1512.00567, 2015. [find](https://scholar.google.com/scholar?q=C+Szegedy%2C+V+Vanhoucke%2C+S+Ioffe%2C+J+Shlens%2C+and+Z+Wojna%2E+Rethinking+the+inception+architecture+for+computer+vision%2E+arXiv+preprint+arXiv%3A1512%2E00567%2C+2015%2E)
26. O Vinyals, S Bengio, and M Kudlur. Order matters: Sequence to sequence for sets. arXiv preprint arXiv:1511.06391, 2015. [find](https://scholar.google.com/scholar?q=O+Vinyals%2C+S+Bengio%2C+and+M+Kudlur%2E+Order+matters%3A+Sequence+to+sequence+for+sets%2E+arXiv+preprint+arXiv%3A1511%2E06391%2C+2015%2E)
27. O Vinyals, M Fortunato, and N Jaitly. Pointer networks. In NIPS, 2015. [find](https://scholar.google.com/scholar?q=O+Vinyals%2C+M+Fortunato%2C+and+N+Jaitly%2E+Pointer+networks%2E+In+NIPS%2C+2015%2E)
28. K Weinberger and L Saul. Distance metric learning for large margin nearest neighbor classification. JMLR, 2009. [find](https://scholar.google.com/scholar?q=K+Weinberger+and+L+Saul%2E+Distance+metric+learning+for+large+margin+nearest+neighbor+classification%2E+JMLR%2C+2009%2E)
29. J Weston, S Chopra, and A Bordes. Memory networks. ICLR, 2014. [find](https://scholar.google.com/scholar?q=J+Weston%2C+S+Chopra%2C+and+A+Bordes%2E+Memory+networks%2E+ICLR%2C+2014%2E)
30. W Zaremba, I Sutskever, and O Vinyals. [Recurrent neural network regularization](https://md.research.elamur.ai/free/en/papers/zaremba-2014-recurrent-neural-network-regularization/). arXiv preprint arXiv:1409.2329, 2014.

*Updated 29 October 2026*