RelationAggregationMethod
Enums · AiDotNet.Enums
Methods for aggregating multiple relation scores in Relation Networks.
For Beginners
In few-shot learning, each class has several example images. When classifying a new query image, we compare it to ALL examples of each class. This enum controls how those multiple similarity scores are combined into a single score for each class.
For example, if we have 5 dog examples and compare a query to each:
- Mean: Average all 5 scores
- Max: Take the highest score (most similar dog example)
- Attention: Weight scores by relevance
- LearnedWeighting: Let the network learn optimal weights
How It Works
When there are multiple support examples per class, we need a way to combine the relation scores from comparing a query with each support example.
Fields
| Field | Summary |
|---|---|
RelationAggregationMethod Attention |
Use attention-weighted average. |
RelationAggregationMethod EmbeddingSum |
Sum each class’s support embeddings and compute one relation per class. |
RelationAggregationMethod LearnedWeighting |
Use learned weighting. |
RelationAggregationMethod Max |
Use maximum score. |
RelationAggregationMethod Mean |
Compute mean of all scores. |
int value__ |