One-Sided Unsupervised Domain Mapping

Sagie Benaim, Lior Wolf

Neural Information Processing Systems 

We present experimental results that the new method not only allows for one sided mapping learning, but also leads to preferable numerical results over the existing circularity-based constraint.