Metric Learning

Proxy-NCA Loss

Variant of Neighborhood Components Analysis that uses learnable proxies to represent each class, reducing computational complexity compared to pair-based approaches. It is particularly effective for large-scale datasets.

← Indietro