Штучний інтелект

Анотація: The application of generative-adversarial networks in the creation of content is studied. Monitoring of training, analysis of architectures, determination of internal processes at the level of layers, research of properties of latent space, and interaction with it are carried out. Variants of using the specified networks in image generation are considered. Special attention is paid to practical implementation aspects, including selecting optimal parameters and data processing. The difference between a classifier and a discriminator is formulated. The principles of generativeadversarial networks and their influence on the efficiency and quality of generated images are studied. The advantages and limitations of using GANs in content creation are considered

Ключові слова: generative adversarial networks, discriminator, generator, image generation, TensorFlow 2

Посилання:

1. Ian J. Goodfellow et al. Generative Adversarial Networks. Advances in Neural Information Processing Systems. 2014. 3(11). Р. 1-9 (DOI: 10.1145/3422622)
2. Course Goggle “Machine Learning”. GAN Variations. URL: https://developers.google.com/machinelearning/gan/applications?hl=ru (Date of application 05.02.2024).
3. Stratehiia rozvytku shtuchnoho intelektu v Ukraini. Za zahalnoiu redaktsiieiu A. I. Shevchenka. Vydavnytstvo «Torpeda». Kyiv. – 2023 r. S 305. (DOI: 10.15407/ development_strategy_2023).
4. Tero Karras, Samuli Laine, Timo Aila. A StyleBased Generator Architecture for Generative Adversarial Networks. Conference: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2019. – P. 4396-4406. (DOI: 10.1109/CVPR.2019.00453)
5. Ian Goodfellow, Yoshua Bengio, Aaron Courville. Deep Learning (Adaptive Computation and Machine Learning series). The MIT Press. 2016. 800р.
6. Andrew Brock, Jeff Donahue, Karen Simonyan. «Large Scale GAN Training for High Fidelity Natural Image Synthesis». Published as a conference paper at ICLR. 2018. Р. 1-35. (https://doi.org/10.48550/arXiv.1809.11096)
7. Pyvovar S.S., Korotka L.I. Mashynne navchannia dlia heneratsii hrafichnykh danykh z vykorystanniam bibliotek TensorFlow ta Keras. Materialy VIII Mizhnarodnoi naukovo-tekhnichnoi konferentsii kompiuterne modeliuvannia ta optymizatsiia skladnykh system (1-3 lystopada 2023 roku m. Dnipro, Ukraina). 2019. S. 128-130.
8. Prykladne mashynne navchannia za dopomohoiu Scikit-Learn ta TensorFlow: kontseptsii, instrumenty ta tekhniky stvorennia intelektualnykh system. Orelen Zheron. Kyiv: «Dyalektyka», 2020. – 688s.
9. Online development service Google Colab. url: https://colab.research.google.com (Date of application 05.02.2024).
10. Klevzhyts D.D., Shvydko D.O., Korotka L.I. Heneratyvno-zmahalni merezhi u sferi stvorennia kontentu. Shtuchnyi intelekt: dosiahnennia, vyklyky ta ryzyky. Mizhnarodna naukova konferentsiia (15-16 bereznia 2024 r., m. Kyiv). 2024. S. 89-94.
11. Liangqu Long, Xiangming Zeng. Beginning Deep Learning with TensorFlow: Work with Keras, MNIST Data Sets, and Advanced Neural Networks. 1st ed. Edition. 2022. 740 р.
12. Aurelien Geron. Hands-On Machine Learning with Scikit-Learn and TensorFlow. O’Reilly Media, Inc. 2017. 684 с.
13. RELU activation function. Official Tensorflow documentation. url: https://www.tensorflow.org/api_docs/python/tf/keras/la yers/ReLU (Date of application 05.02.2024).
14. David Foster. Generative Deep Learning. O’Reilly Media. 2019. 327 с.