Generalizes the Variational Autoencoder using the methods of Nonlinear Statistical Coupling a. The Coupled Evidence Lower Bound uses generalized logarithms to provide tuning of the loss functions, particularly to improve the robustness of the learned models. b. The Coupled Normal Distribution generalizes the latent layer providing increased flexibility regarding the learned model.
hxyue1 / coupled-vae Goto Github PK
View Code? Open in Web Editor NEWThis project forked from photrek/coupled-vae
Machine learning algorithms for complex systems leveraging the methods of nonlinear statistical coupling
License: GNU General Public License v3.0