Authors : Maxime Berillon and Rémy Deshayes
Generative modeling is an unsupervised learning task that aims at discovering and learning the underlying
structure of an input data in such a way that the model can be used to generate new examples that plausibly
could have been drawn from the input’s dataset.
As always with unsupervised learning our main task will amount to learning the probability distribution of
that data at hand.
Section 1 of our report - please have a look at it - introduces one method to learn a probability distribution and the GANs a very famous generative method leveraging this technique. Then, section 2 discusses the concept of distance between probability distributions and presents a very famous distance taken from the optimal transport theory : the Wasserstein distance. As an alternative to the WGAN, section 2 also compares the primal and dual approach of the optimal transport problem introduced. Building on those remarks, section 3, introduces the OT-GAN and its implementation.