WeproposetheWassersteinAuto-Encoder(WAE)|anewalgorithmforbuildingagenerativemodelofthedatadistribution.WAEminimizesapenalizedformoftheWassersteindistancebetweenthemodeldistributionandthetargetdistribution,whichleadstoadierentregularizerthantheoneusedbytheVariationalAuto-Encoder(VAE)[1