Módszerek összehasonlítása

	Transzfer tanulás	Few-shot Learning	Félfelügyelt tanulás
Tudományterület	Gépi tanulás	Gépi tanulás	Gépi tanulás
Módszercsalád	Machine learning	Machine learning	Machine learning
Keletkezés éve	2010 (formalized); 1990s (early roots)	2011–2017	1970s–2006 (formalized)
Megalkotó	Pan, S. J. & Yang, Q. (survey); Bengio, Y. (deep learning framing)	Lake, B. M.; Vinyals, O.; Finn, C. et al.	Vapnik, V. N. and others (community of researchers, 1970s–2000s)
Típus	Learning paradigm	Meta-learning / low-data learning paradigm	Learning paradigm
Alapmű	Pan, S. J., & Yang, Q. (2010). A Survey on Transfer Learning. IEEE Transactions on Knowledge and Data Engineering, 22(10), 1345–1359.	Vinyals, O., Blundell, C., Lillicrap, T., Wierstra, D., & Kavukcuoglu, K. (2016). Matching Networks for One Shot Learning. Advances in Neural Information Processing Systems (NeurIPS), 29.	Chapelle, O., Scholkopf, B., & Zien, A. (Eds.) (2006). Semi-Supervised Learning. MIT Press. ISBN: 978-0-262-03358-9
Alternatív nevek	TL, domain adaptation, fine-tuning, pre-trained model adaptation	FSL, low-shot learning, k-shot learning, meta-learning for few examples	SSL, semi-supervised machine learning, transductive learning, label-efficient learning
Kapcsolódó	3	4	5
Összefoglaló	Transfer learning is a machine learning paradigm in which knowledge gained from training a model on a source task or domain is reused to improve learning on a different but related target task or domain. It is especially powerful when labeled data for the target task is scarce, and it underlies most modern deep learning applications in computer vision, natural language processing, and beyond.	Few-shot learning is a machine learning paradigm that trains models to recognize new classes or solve new tasks from only a handful of labeled examples — typically one to five — by leveraging prior knowledge acquired from a large, related training distribution. It is especially relevant in domains where labeling is expensive, scarce, or structurally limited.	Semi-supervised learning (SSL) is a machine learning paradigm that trains models using a small set of labeled examples together with a much larger pool of unlabeled data. By leveraging the structure inherent in unlabeled data, SSL achieves accuracy closer to fully supervised models while requiring far fewer costly manual labels — making it practical when labeling is expensive, slow, or resource-constrained.