Principal Bit Analysis:
Autoencoding with Schur-Concave Loss
Sourbh Bhadane 1 Aaron B. Wagner 1 Jayadev Acharya 1
Abstract and g are selected through training from the class of func-
tions realized by multilayer perceptrons of a given architec-
We consider a linear autoencoder in which the
ture (Hinton & Salakhutdinov, 2006). Yet, the canonical
latent variables ...


雷达卡




京公网安备 11010802022788号







