Evaluate "tuplet loss" - Githubissues

hominot / metric_learning

0 stars 0 forks source link

Evaluate "tuplet loss" #17

Closed hominot closed 5 years ago

hominot commented 5 years ago

Hyperparameters:

Parametrization (Euclidean, dot product)
L2 regularizing parameter
(N+1)-pair size (N=2, 71, 191 ...)
Learning rate