Groups
Category
Triplet loss and contrastive loss are metric-learning objectives that teach a model to map similar items close together and dissimilar items far apart in an embedding space.
Generalization bounds explain why deep neural networks can perform well on unseen data despite having many parameters.