Metrix

2000년 도입 · 논문 1편에서 사용

A generic way of representing and interpolating labels, which allows straightforward extension of any kind of mixup to deep metric learning for a large class of loss functions.

출처: It Takes Two to Tango: Mixup for Deep Metric Learning

소개 논문: It Takes Two to Tango: Mixup for Deep Metric Learning