Few-shot learning

Few-shot learning (FSL) is a problem setup in machine learning in which a model learns to perform a task, typically classification, from only a small number of labeled examples per class, rather than the large datasets required by conventional supervised learning.

One-shot learning is the special case of the N-way K-shot framing in which K equals one, such that the model must generalize from exactly one example per class.

The limiting case of few-shot learning is zero-shot learning that requires no training on examples of the class to be classified; instead, the classification task must be generalized from examples of other classes.

Applications

Few-shot learning is useful in many real-life applications of artificial intelligence. It allows computers to learn new things even when only a few examples are available. For example, in image recognition, a computer can learn to identify a new object by studying a small number of pictures. In language processing, it helps AI systems understand and respond to tasks with only a few examples. It can also be helpful in healthcare, where AI models can study medical images and support the identification of diseases. This approach is useful when collecting large amounts of data is difficult or takes a lot of time.

See also

References

  1. 1 2 Wang, Yaqing; Yao, Quanming; Kwok, James T.; Ni, Lionel M. (2020-06-12). "Generalizing from a Few Examples: A Survey on Few-shot Learning". ACM Comput. Surv. 53 (3): 63:1–63:34. doi:10.1145/3386252. ISSN 0360-0300.
  2. ↑ Song, Yisheng; Wang, Ting; Cai, Puyu; Mondal, Subrota K.; Sahoo, Jyoti Prakash (2023-07-13). "A Comprehensive Survey of Few-shot Learning: Evolution, Applications, Challenges, and Opportunities". ACM Comput. Surv. 55 (13s): 271:1–271:40. doi:10.1145/3582688. ISSN 0360-0300.
  3. ↑ Parnami, Archit; Lee, Minwoo (2022-03-07), Learning from Few Examples: A Summary of Approaches to Few-Shot Learning, arXiv:2203.04291
  4. ↑ Han, Xiaoming; Shi, Dianxi; Wang, Zhen; Chen, Yang; Jin, Songchang; Yang, Shaowu (2026-01-17). "Meta-learning for few-shot open task recognition". Scientific Reports. 16 (1): 5624. Bibcode:2026NatSR..16.5624H. doi:10.1038/s41598-026-36291-x. ISSN 2045-2322. PMC 12891690. PMID 41547899.
  5. ↑ Tyukin, Ivan Y.; Gorban, Alexander N.; Alkhudaydi, Muhammad H.; Zhou, Qinghua (2021-07-18). "Demystification of Few-shot and One-shot Learning". 2021 International Joint Conference on Neural Networks (IJCNN). IEEE. pp. 1–7. Bibcode:2021ijcn.conf..800T. doi:10.1109/IJCNN52387.2021.9534395. ISBN 978-1-6654-3900-8.
  6. 1 2 Kadam, Suvarna; Vaidya, Vinay (2020). "Review and Analysis of Zero, One and Few Shot Learning Approaches". In Abraham, Ajith; Cherukuri, Aswani Kumar; Melin, Patricia; Gandhi, Niketa (eds.). Intelligent Systems Design and Applications. Advances in Intelligent Systems and Computing. Vol. 940. Cham: Springer International Publishing. pp. 100–112. doi:10.1007/978-3-030-16657-1_10. ISBN 978-3-030-16657-1.
  7. ↑ Vinyals, Oriol; Blundell, Charles; Lillicrap, Timothy; Kavukcuoglu, Koray; Wierstra, Daan (2016-12-05). Matching networks for one shot learning. Red Hook, NY, USA: Curran Associates Inc. pp. 3637–3645. ISBN 978-1-5108-3881-9.
  8. ↑ Wang, Wei; Zheng, Vincent W.; Yu, Han; Miao, Chunyan (2019-01-16). "A Survey of Zero-Shot Learning: Settings, Methods, and Applications". ACM Trans. Intell. Syst. Technol. 10 (2): 13:1–13:37. doi:10.1145/3293318. ISSN 2157-6904.
  9. ↑ Xian, Yongqin; Lampert, Christoph H.; Schiele, Bernt; Akata, Zeynep (2019-09-01). "Zero-Shot Learning—A Comprehensive Evaluation of the Good, the Bad and the Ugly". IEEE Transactions on Pattern Analysis and Machine Intelligence. 41 (9): 2251–2265. Bibcode:2019ITPAM..41.2251X. doi:10.1109/TPAMI.2018.2857768. ISSN 0162-8828. PMID 30028691.
  10. ↑ Pachetti, Eva; Colantonio, Sara (2024-10). "A systematic review of few-shot learning in medical imaging". Artificial Intelligence in Medicine. 156: 102949. doi:10.1016/j.artmed.2024.102949. ISSN 1873-2860. PMID 39178621. {{cite journal}}: Check date values in: |date= (help)CS1 maint: article number as page number (link)