Generative Adversarial Networks

from Wikipedia, the free encyclopedia

Generative Adversarial Networks (GAN, in German for 'generating opposing networks') are a group of algorithms for unsupervised learning in computer science .

properties

Generative Adversarial Networks consist of two artificial neural networks that perform a zero-sum game . One of them creates candidates (the generator), the second neural network evaluates the candidates (the discriminator ). The generator typically maps a vector of latent variables to the desired result space. The goal of the generator is to learn to generate results according to a certain distribution . The discriminator, on the other hand, is trained to distinguish the results of the generator from the data from the real, given distribution. The generator's objective function is then to produce results that the discriminator cannot distinguish. As a result, the generated distribution should gradually adjust to the real distribution.

use

GANs have been used, among other things, to create photo-realistic images to visualize various objects, to model movement patterns in videos, to create 3D models of objects from 2D images, and to manipulate astronomical images. GANs are also used to naturally shape user interaction with chatbots. GANs are also used in particle physics to accelerate time-consuming detector simulations.

history

The use of competitive neural networks was first proposed in 2013 by Wei Li, Melvin Gauci, and Roderich Gross. The concept of Generative Adversarial Networks was developed in 2014 by Yoshua Bengio , Ian Goodfellow , Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair and Aaron Courville. Following the Turing test method is also called "Turing Learning" ( English Turing learning refers).

literature

  • Ian Goodfellow, Yoshua Bengio and Aaron Courville: Deep Learning (Adaptive Computation and Machine Learning) , MIT Press, Cambridge (USA), 2016. ISBN 978-0262035613 .

Individual evidence

  1. ^ A b Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio: Generative Adversarial Networks . In: NIPS . 2014.
  2. ↑ Generating Videos with Scene Dynamics .
  3. ^ 3D Generative Adversarial Network .
  4. ↑ Kevin Schawinski, Ce Zhang, Hantian Zhang, Lucas Fowler, Gokula Krishnan Santhanam: Generative Adversarial Networks recover features in astrophysical images of galaxies beyond the deconvolution limit . In: Instrumentation and Methods for Astrophysics . February 1, 2017. arxiv : 1702.00403 .
  5. ↑ Larry Greenemeier: When Will Computers Have Common Sense? Ask Facebook . June 20, 2016. Accessed July 31, 2016.
  6. ↑ Michela Paganini, Luke de Oliveira, Benjamin Nachman: Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis . In: Computing and Software for Big Science . 1, 2017, p. 4. arxiv : 1701.05927 . bibcode : 2017arXiv170105927D . doi : 10.1007 / s41781-017-0004-6 .
  7. ↑ Martin Erdmann, Jonas Glombitza, Thorben Quast: Precise Simulation of Electromagnetic Calorimeter Showers Using a Wasserstein Generative Adversarial Network . In: Computing and Software for Big Science . 3, 2019, p. 4. arxiv : 1807.01954 . doi : 10.1007 / s41781-018-0019-7 .
  8. ^ Wei Li, Melvin Gauci, and Roderich Gross: A Coevolutionary Approach to Learn Animal Behavior Through Controlled Interaction. In: Proceedings of the 15th Annual Conference on Genetic and Evolutionary Computation (GECCO 2013) Amsterdam, July 6, 2013, pp. 223-230. doi : 10.1145 / 2463372.2465801
  9. ^ Wei Li, Melvin Gauci, Roderich Groß: Turing learning: a metric-free approach to inferring behavior and its application to swarms . In: Swarm Intelligence . 10, No. 3, August 30, 2016, pp. 211–243. doi : 10.1007 / s11721-016-0126-1 .