Hands-On One-shot Learning with Python

Triplet loss function

The triplet loss function is an alternative to the contrastive loss function. It has convergence advantages over contrastive loss functions.

To learn about the triplet loss function, first, we need to define data points in pairs as follows:

• Anchor (A): The main data point
• Positive (P): A data point similar to Anchor
• Negative (N): A different data point than Anchor

Considering is the output of Siamese networks, ideally, we can assume the following:

In distance function terms, we can say the following:

As we don't want ...