方法对比

	课程学习	知识蒸馏	迁移学习
领域	深度学习	深度学习	机器学习
方法族	Machine learning	Machine learning	Machine learning
起源年份	2009	2015	2010 (formalized); 1990s (early roots)
提出者	Yoshua Bengio et al.	Hinton, G., Vinyals, O. & Dean, J.	Pan, S. J. & Yang, Q. (survey); Bengio, Y. (deep learning framing)
类型	Training strategy	Neural network compression (teacher–student)	Learning paradigm
开创性文献	Bengio, Y., Louradour, J., Collobert, R., & Weston, J. (2009). Curriculum learning. International Conference on Machine Learning (ICML), 41–48.	Hinton, G., Vinyals, O. & Dean, J. (2015). Distilling the Knowledge in a Neural Network. NeurIPS Deep Learning Workshop.	Pan, S. J., & Yang, Q. (2010). A Survey on Transfer Learning. IEEE Transactions on Knowledge and Data Engineering, 22(10), 1345–1359.
别名	Scheduled Training, Difficulty-Based Training, Self-Paced Learning, Müfredat Öğrenimi	Bilgi Damıtma (Knowledge Distillation), bilgi damıtma, teacher-student distillation, model distillation	TL, domain adaptation, fine-tuning, pre-trained model adaptation
相关	3	5	3
摘要	Curriculum Learning is a training strategy for machine learning models, introduced by Bengio et al. in 2009, in which training examples are presented in a meaningful order—typically from easy to hard—rather than at random. Inspired by how humans and animals learn progressively, it organizes training data into a curriculum that starts with simpler, cleaner, or more representative samples and gradually introduces harder or more complex examples as the model matures.	Knowledge Distillation is a model-compression technique, introduced by Geoffrey Hinton and colleagues in 2015, that trains a small student model using the soft-label outputs of a large teacher model. Distilled models such as DistilBERT and TinyBERT reach roughly 97% of the larger model's performance while running far faster.	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.