AI RESEARCH
Curved representational Bregman divergences and their applications
arXiv CS.LG
•
ArXi:2504.05654v5 Announce Type: replace-cross By analogy to the terminology of curved exponential families in statistics, we define curved Bregman divergences as Bregman divergences restricted to non-affine parameter subspaces and sub-dimensional Bregman divergences when the restrictions are affine. A common example of curved Bregman divergence is the cosine dissimilarity between normalized vectors: a curved squared Euclidean divergence.