Salīdzināt metodes

	Pašuzraudzītā aktīvā mācīšanās	Pašuzraudzības apmācība
Nozare	Mašīnmācīšanās	Mašīnmācīšanās
Saime	Machine learning	Machine learning
Izcelsmes gads	2020–2021	2018–2020
Autors	Bengar et al. and concurrent works (multiple groups)	LeCun, Y. and community (formalized ~2018–2020)
Tips	Hybrid active-learning and self-supervised pre-training framework	Representation learning paradigm
Pirmavots	Bengar, J. Z., van de Weijer, J., Twardowski, B., & Raducanu, B. (2021). Reducing Label Effort: Self-Supervised Meets Active Learning. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), pp. 1631–1639.	LeCun, Y. & Misra, I. (2022). Self-supervised learning: The dark matter of intelligence. Meta AI Blog. https://ai.facebook.com/blog/self-supervised-learning-the-dark-matter-of-intelligence/
Citi nosaukumi	SSL-AL, self-supervised active learning, semi-supervised active learning with self-supervision, label-efficient self-supervised learning	SSL, self-supervised pre-training, pretext-task learning, unsupervised representation learning
Saistītās	5	3
Kopsavilkums	Self-supervised Active Learning (SSL-AL) is a label-efficient machine-learning paradigm that pre-trains a model on unlabeled data using self-supervised objectives, then strategically queries a human oracle for the most informative labels using an active-learning acquisition function. The result is strong predictive performance with a fraction of the annotation cost required by fully supervised approaches.	Self-supervised learning (SSL) is a machine-learning paradigm that generates its own supervisory signal directly from unlabeled data by defining an auxiliary pretext task — such as predicting masked words, rotating images, or contrasting augmented views — and uses the learned representations as a powerful starting point for downstream tasks with minimal labeled examples.