Generative Adversarial Neural Network for Creating Photorealistic Images

This paper is focused on studying the Generative Adversarial Neural Network (GAN or GANN) as an implement for creating diverse functional samples, particularly photorealistic images (graphic, molecular, etc.). The paper considers available existing methods and approaches for designing and algorithmization the current class of networks, also the effectiveness of different types of formed architectures with various combinations using the example of handwritten digits creation as one of the photorealistic images. The paper examines an applied value of the generative adversarial neural network as an implementation of the complex paradigm of artificial intelligence. The results of the study demonstrate the efficiency of the GAN technology in designing samples of various types and categories of complexity