Investigation of generally-standing network

Authors: Ayrapetov A.E., Kovalenko A.A.
Published in issue: #10(27)/2018
DOI: 10.18698/2541-8009-2018-10-380
Category: Informatics, Computer Engineering and Control | Chapter: System Analysis, Control, and Information Processing, Statistics
Keywords: generative-adversarial network, machine learning, neural networks, training without a teacher, generator, discriminator, perceptron, autocoder
Published: 01.10.2018

The article deals with the algorithms of generative-adversarial neural networks, which are relatively young and one of the most promising in the family of machine learning algorithms without a teacher. Currently, such neural networks are used to generate photorealistic images based on the examples provided to it. However, the potential of this subgroup of algorithms is not fully disclosed and allows increasing the plausibility of generated images. The paper describes the features of the structure of the generative-adversarial neural network, analyzes its accuracy and performance, and also provides solutions that allow optimizing the existing learning algorithm and the topology of the neural network.

References

[1] GAN — generativnye sostyazatel’nye seti [GAN — generative adversarial networks]. Available at: http://robocraft.ru/blog/machinelearning/3693.html (accessed 10 May 2018).

[2] GAN: a beginner’s guide to generative adversarial networks. Available at: https://deeplearning4j.org/generative-adversarial-network (accessed 15 May 2018).

[3] Goodfellow I.J., Pouget-Abadie J., Mirza M., Xu B., Warde-Farley D., Ozair Sh., Courville A., Bengio Yo. Generative adversarial networks. Universite de Montreal, 2014, 9 p.

[4] Generative adversarial networks. Available at: https://habr.com/post/352794 (accessed 17 May 2018).

[5] Ganin Ya., Lempitsky V. Unsupervised domain adaptation by backpropagation. Available at: https://arxiv.org/pdf/1409.7495.pdf (accessed 15 May 2018).

[6] Gitman I. Adversarial networks. Available at: http://www.machinelearning.ru/wiki/images/4/4f/Gan.pdf (accessed 20 May 2018).

[7] How to train a GAN? Tips and tricks to make GANs work. Available at: https://github.com/soumith/ganhacks (accessed 21 May 2018).

[8] Mirza M., Osindero S. Conditional generative adversarial nets. Available at: https://arxiv.org/pdf/1411.1784.pdf (accessed 15 May 2018).

[9] A beginner’s guide to deep autoencoders. Available at: https://deeplearning4j.org/deepautoencoder (accessed 28 May 2018).

[10] Makhzani A., Shlens J., Jaitly N., Goodfellow I., Frey B. Adversarial autoencoders. Available at: https://arxiv.org/pdf/1511.05644.pdf (accessed 15 May 2018).