Sammenlign metoder
Gjennomgå de valgte metodene side om side; rader som avviker, er uthevet.
| Ensemble Self-supervised Learning× | Kunnskapsdestillasjon× | |
|---|---|---|
| Fagfelt≠ | Maskinlæring | Dyp læring |
| Familie | Machine learning | Machine learning |
| Opprinnelsesår≠ | 2020–2021 | 2015 |
| Opphavsperson≠ | Multiple contributors (Grill et al., Caron et al., Chen et al.) | Hinton, G., Vinyals, O. & Dean, J. |
| Type≠ | Ensemble of self-supervised models or objectives | Neural network compression (teacher–student) |
| Opprinnelig kilde≠ | Grill, J.-B., Strub, F., Altché, F., Tallec, C., Richemond, P. H., Buchatskaya, E., Doersch, C., Ávila Pires, B., Guo, Z., Gheshlaghi Azar, M., Piot, B., Kavukcuoglu, K., Munos, R., & Valko, M. (2020). Bootstrap Your Own Latent: A New Approach to Self-Supervised Learning. Advances in Neural Information Processing Systems, 33, 21271–21284. link ↗ | Hinton, G., Vinyals, O. & Dean, J. (2015). Distilling the Knowledge in a Neural Network. NeurIPS Deep Learning Workshop. link ↗ |
| Alias | ensemble SSL, multi-view self-supervised ensemble, self-supervised ensemble learning, SSL ensemble | Bilgi Damıtma (Knowledge Distillation), bilgi damıtma, teacher-student distillation, model distillation |
| Relaterte | 5 | 5 |
| Sammendrag≠ | Ensemble Self-supervised Learning combines multiple self-supervised models, objectives, or augmentation views into a unified framework to produce more robust and generalizable representations from unlabeled data. By aggregating diverse self-supervised signals, the ensemble reduces the risk of representation collapse and outperforms single-objective SSL approaches on downstream tasks. | 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. |
| ScholarGateDatasett ↗ |
|
|