GO Loss: A Gaussian Distribution-Based Orthogonal Decomposition Loss for Classification

Table 1

Classification accuracy on MNIST dataset.

Methods	Remark	Acc. (%)
Soft-max	—	99.28
Center loss [27]	λ = 0.1	99.62
Ring loss [23]	λ = 0.1	99.58
LGM loss [28]	α = 1	99.36
GCPL loss [37]	λ = 0.1	99.41
SL [41]		99.32
GO loss		99.660.03