Generative Networks

The course aims for students to gain deep knowledge of modern theoretical methods and technological implementations linked to generative methodologies. With the help of the necessary software and hardware device systems, students learn about the theoretical and practical backgrounds of the components of generative methodologies, basic generative networks in the PyTorch environment, DCGAN that also uses advanced convolution layers, GAN control, and the creation of conditional GANs. With the help of the course, students learn about advanced GAN programming in a practice-oriented way, such as data augmentation, protection of personal data, or the use of GAN applications in questionnaires. The course pays special attention to complex solutions, including measuring the comparability of generative models, realism and diversification, the detection of bias, or implementing different style transfer techniques (Pix2Pix, CycleGAN, StyleGAN). Students work on pre-agreed project tasks within the areas of generative methodology applications.

======

Recommended Literatures and Courses

Jakub Langr and Vladimir Bok: GANs in Action, Manning, 2019. David Foster: Generative Deep Learning, Oreilly, 2019. Kailash Ahirwar: Generative Adversarial Networks Projects, Packt, 2019. I. Goodfellow, Y. Bengio, A. Courville: Deep Learning, MIT Press, 2016. Stuart Russell and Peter Norvig: Artificial Intelligence: A Modern Approach, 4th US ed., Pearson, 2020.