contrastive
7.8
Relating to methods that train models by comparing similar and different examples
- adjective
- /kənˈtræstɪv/
- Specialized
:
contrastivo
- In machine learning, contrastive loss is crucial for training models with dissimilar data pairs.
But that said, there's non contrastive methods, right?
- But that said, there's non contrastive methods, right?
Examples
-
That's the idea and it's and we call this process clip right or contrastive language image pairs.
-
The researchers used a contrastive method to improve the accuracy of their predictions.
-
Contrastive learning helps models understand the differences between contrastive images.
Surface Forms
contrastive positive
Morphology
Etymology
Contrastive comes from the root contra- meaning 'against', and in machine learning this idea stays the same: the method puts items 'against' each other to see which are similar and which are different. So a contrastive approach trains by comparing pairs to teach the model what is alike and what is not.
