We analyze in detail the interactions between the models that make up the GANs, which allows us to make modifications to adapt these architectures to specific use cases. The comprehension of the interplay between generator and discriminator not only facilitates an optimal use of these networks, but also leads to the proposal of new types of interactions between models, which may be the next step beyond GANs.

Development of algorithms with new architectures in NLP that are competitive with current architectures. The research focuses on the design of new architectures, layers, error functions and optimization, which together define a model centered on text analysis for image generation (Text2Image).