Definition
Similarity learning focuses on training models to map inputs into a vector space where similar items are close together and dissimilar items are far apart. Techniques like Siamese networks or triplet loss are commonly used. Instead of predicting explicit labels, the model learns a representation that preserves semantic relationships, enabling efficient retrieval, verification, and clustering tasks by comparing distances in the embedding space rather than relying on direct classification boundaries.
Summary
A machine learning approach that learns a distance metric to determine how similar or dissimilar two objects are.
Key Concepts
- Embedding space
- Distance metrics
- Triplet loss
- Siamese networks
Use Cases
- Face recognition
- Duplicate detection
- Recommendation systems