Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?
Nenad TomasevIoana BicaBrian McWilliamsLars BuesingRazvan PascanuCharles BlundellJovana MitrovicPublished in: CoRR (2022)
Keyphrases
- supervised learning
- class labels
- training examples
- training data
- training set
- unsupervised learning
- semi supervised
- labeled examples
- supervised classification
- object categories
- active learning
- semi supervised learning
- learning algorithm
- pairwise
- labeled data
- learning problems
- image collections
- partially labeled data
- supervised machine learning
- unlabeled data
- machine learning
- reinforcement learning
- object classes
- statistical learning
- multi label
- hierarchical structure
- label propagation
- neural network
- image labeling
- computer vision
- multiple instance learning
- decision trees
- generalization error
- image data
- d objects
- small number
- markov random field