Triplet Loss

Triplet Loss uses an anchor, a positive example that is similar to the anchor, and a negative example that is different from it. During training, the model learns to reduce the distance between the anchor and positive example while increasing the distance between the anchor and negative example, usually by a specified margin.

Triplet Loss helps models learn meaningful feature representations and distinguish between similar and dissimilar data points. This makes it useful when the goal is to compare or match examples based on their learned representations rather than simply assign them to predefined classes.