> ML_LITERATURE // CHEN-2020-SIMPLE-FRAMEWORK-CONTRASTIVE-LEARNING-VISUAL-REPRESENTATIONS-SIMCLR_v1.0
A Simple Framework for Contrastive Learning of Visual Representations (SimCLR)
Ting Chen, Simon Kornblith, Mohammad Norouzi, Geoffrey Hinton · International Conference on Machine Learning (ICML) (2020)
algorithm2020foundationalthirdPartyReproduced
Principal Contribution
Showed that simple contrastive learning with strong composite data augmentations, non-linear projection heads, and large batch sizes matches supervised ImageNet pre-training.
Operational Relevance
Serves as qualified reference for deploying task-feature-extraction, task-image-classification in production.
Assumptions
- Spatial feature coherence and data manifold structure adhere to continuous representation hypotheses
Limitations
- Computational complexity scales with spatial resolution and parameter capacity
Connected Algorithms, Architectures & Tools
Related Algorithms:
Related Architectures:
Implementing Libraries: