TripletLossFitnessCalculator<T, TInput, TOutput>
Models & Types · AiDotNet.FitnessCalculators
A fitness calculator that uses Triplet Loss to evaluate model performance, particularly for similarity learning and embedding tasks.
Inheritance: FitnessCalculatorBase<T, TInput, TOutput> → TripletLossFitnessCalculator<T, TInput, TOutput>
For Beginners
This calculator helps train models that learn to recognize similarities between items.
Triplet Loss works with three examples at a time:
- An “anchor” (the reference item)
- A “positive” (an item similar to the anchor)
- A “negative” (an item different from the anchor)
The goal is to teach the model to:
- Push the anchor and positive close together in the feature space
- Push the anchor and negative far apart in the feature space
Think of it like organizing a bookshelf:
- You have a fantasy novel (anchor)
- You want to place other fantasy novels (positives) close to it
- You want to place non-fantasy books (negatives) far away from it
Triplet Loss is especially useful for:
- Face recognition (same person = positive, different person = negative)
- Product recommendations (similar products = positive, different products = negative)
- Image search (visually similar images = positive, different images = negative)
- Document similarity (documents on same topic = positive, different topics = negative)
The “margin” parameter controls how far apart the negative examples should be from the anchor compared to the positive examples.
Constructors
| Constructor | Summary |
|---|---|
TripletLossFitnessCalculator(T margin, DataSetType dataSetType) |
Initializes a new instance of the TripletLossFitnessCalculator class. |