Wasserstein Introspective Neural Networks

11/24/2017
by   Kwonjoon Lee, et al.
0

We present Wasserstein introspective neural networks (WINN) that are both a generator and a discriminator within a single model. WINN provides a significant improvement over the recent introspective neural networks (INN) method by enhancing INN's generative modeling capability. WINN has three interesting properties: (1) A mathematical connection between the formulation of Wasserstein generative adversarial networks (WGAN) and the INN algorithm is made; (2) The explicit adoption of the Wasserstein distance into INN results in a large enhancement to INN, achieving compelling results even with a single classifier --- e.g., providing a 20 times reduction in model size over INN within texture modeling; (3) When applied to supervised classification, WINN also gives rise to greater robustness with an 88% reduction of errors against adversarial examples --- improved over the result of 39% by an INN-family algorithm. In the experiments, we report encouraging results on unsupervised learning problems including texture, face, and object modeling, as well as a supervised classification task against adversarial attack.

READ FULL TEXT

Please sign up or login with your details

Forgot password? Click here to reset