Generative-Adversarial Networks

Brief course description This discipline is devoted to the latest methods of generative-adversarial networks, and their use to create realistic images and three-dimensional structures. Upon mastering the discipline students should know: the concept and organisation of the generative model; the concept and organisation of the discriminative model; be able to: train generative-adversarial networks and generate images using them, from basic handwritten digits, to restoration, correction, colouring of photographs; generate 3D. Expected learning outcomes Upon successful completion of the course, students will be able to: - Possess an advanced understanding of the principles of generative-adversarial networks, their structure and basic components. - Design and train generative-adversarial models for different types of data such as images, text, sound and others. - apply GSS in various domains such as computer vision, natural language processing, content generation and other creative applications. - tune and optimise GSS parameters to achieve better results in various tasks.