Variance Covariance Regularization Enforces Pairwise Independence in Self-Supervised Representations.
Grégoire MialonRandall BalestrieroYann LeCunPublished in: CoRR (2022)
Keyphrases
- pairwise
- covariance matrix
- higher order
- markov random field
- loss function
- multi class
- square loss
- similarity measure
- semi supervised
- high order
- pairwise interactions
- belief propagation
- potential functions
- regularization term
- standard deviation
- graph matching
- symbolic representation
- similarity function
- parameter selection
- pairwise constraints
- machine learning
- decision trees