Generative Adversarial Networks

Generative Competitive Networks (GANs) is a promising technology that is widely regarded as one of the most interesting achievements in AI and deep learning over the past decade. First introduced in a groundbreaking research paper by Ian Goodfellow et al. in 2014, this model has become the focus of conversation in academic and industrial communities due to its potential applicability in a wide range of problem areas, in particular, in image-to-image translation tasks. GANs are a powerful framework for training artificial neural networks in an unsupervised competitive mode. This study has two objectives: to explore the idea of generative competitive modeling for unsupervised learning tasks and to analyze the latent representation spaces obtained by such models. Several experiments on changing various sensible characteristics of images have been conducted, and a prototype allowing to repeat the experiments has been developed.