Curriculum learning¶
Bengio, Y., Louradour, J., Collobert, R., & Weston, J. (2009). Curriculum learning. Proceedings of the 26th Annual International Conference on Machine Learning (ICML '09), 41-48.
Cited by¶
2 citations across 2 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Pedagogy
- The pattern survives the removal of the schoolroom, the human teacher, and the conscious learner, which is what licenses calling it a prime rather than an education-domain specialty, as Bengio et al. (2009) demonstrated when their curriculum-learning result imported the easy-to-hard sequencing move directly into neural-network training without any human in the loop.
This sourceDemonstrates that ordering training examples easy-to-hard improves neural-network generalization, importing the pedagogical sequencing move into ML without a human in the loop — the load-bearing substrate-independence demonstration this prime cites.
- The pattern survives the removal of the schoolroom, the human teacher, and the conscious learner, which is what licenses calling it a prime rather than an education-domain specialty, as Bengio et al. (2009) demonstrated when their curriculum-learning result imported the easy-to-hard sequencing move directly into neural-network training without any human in the loop.
- Zone of Proximal Development (ZPD)
- The same ordered-difficulty intuition is formalized in machine-learning curriculum learning by Bengio, Louradour, Collobert, and Weston (2009), where models trained on progressively harder examples converge faster and to better optima — an algorithmic homologue of ZPD-targeted instructional sequencing.
This sourceACM. Introduces curriculum learning in machine learning: training models on progressively harder examples improves convergence and final performance — algorithmic homologue of ZPD-targeted instructional sequencing.
- The same ordered-difficulty intuition is formalized in machine-learning curriculum learning by Bengio, Louradour, Collobert, and Weston (2009), where models trained on progressively harder examples converge faster and to better optima — an algorithmic homologue of ZPD-targeted instructional sequencing.
Verification¶
This reference passed the adversarial substantiation pipeline: it was checked to exist and to support the claim it is attached to. See how references were verified.
Registry ID ref:cb06cd6fdba9 · see in the full table