Ecological Analogy for Generative Adversarial Networks and Diversity Control
Generative adversarial networks are popular deep neural networks for generative modeling in the field of artificial intelligence. In the generative modeling, we want to output a sample with some random numbers as an input. We train the artificial neural network with a training data set for the purpo...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2022-01-01
|
| Series: | Journal of Physics: Complexity |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2632-072X/acacdf |
| _version_ | 1797844605416243200 |
|---|---|
| author | Kenichi Nakazato |
| author_facet | Kenichi Nakazato |
| author_sort | Kenichi Nakazato |
| collection | DOAJ |
| description | Generative adversarial networks are popular deep neural networks for generative modeling in the field of artificial intelligence. In the generative modeling, we want to output a sample with some random numbers as an input. We train the artificial neural network with a training data set for the purpose. The network is known with astonishingly fruitful demonstrations, but we know the difficulty in the training because of the complex training dynamics. Here, we introduce an ecological analogy for the training dynamics. With the simple ecological model, we can understand the dynamics. Furthermore, a controller for the training can be designed based on the understanding. We then demonstrate how the network and the controller work with an ideal case, MNIST. |
| first_indexed | 2024-04-09T17:26:00Z |
| format | Article |
| id | doaj.art-6ebe5a8c115e4edcbd6113ea65699749 |
| institution | Directory Open Access Journal |
| issn | 2632-072X |
| language | English |
| last_indexed | 2024-04-09T17:26:00Z |
| publishDate | 2022-01-01 |
| publisher | IOP Publishing |
| record_format | Article |
| series | Journal of Physics: Complexity |
| spelling | doaj.art-6ebe5a8c115e4edcbd6113ea656997492023-04-18T13:50:38ZengIOP PublishingJournal of Physics: Complexity2632-072X2022-01-014101LT0110.1088/2632-072X/acacdfEcological Analogy for Generative Adversarial Networks and Diversity ControlKenichi Nakazato0https://orcid.org/0000-0001-5001-1005Bosch Center for Artificial Intelligence, Robert Bosch Japan , Tokyo, JapanGenerative adversarial networks are popular deep neural networks for generative modeling in the field of artificial intelligence. In the generative modeling, we want to output a sample with some random numbers as an input. We train the artificial neural network with a training data set for the purpose. The network is known with astonishingly fruitful demonstrations, but we know the difficulty in the training because of the complex training dynamics. Here, we introduce an ecological analogy for the training dynamics. With the simple ecological model, we can understand the dynamics. Furthermore, a controller for the training can be designed based on the understanding. We then demonstrate how the network and the controller work with an ideal case, MNIST.https://doi.org/10.1088/2632-072X/acacdfdeep neural networksecological dynamicsgenerative adversarial networkstraining dynamicsbio-inspired algorithm |
| spellingShingle | Kenichi Nakazato Ecological Analogy for Generative Adversarial Networks and Diversity Control Journal of Physics: Complexity deep neural networks ecological dynamics generative adversarial networks training dynamics bio-inspired algorithm |
| title | Ecological Analogy for Generative Adversarial Networks and Diversity Control |
| title_full | Ecological Analogy for Generative Adversarial Networks and Diversity Control |
| title_fullStr | Ecological Analogy for Generative Adversarial Networks and Diversity Control |
| title_full_unstemmed | Ecological Analogy for Generative Adversarial Networks and Diversity Control |
| title_short | Ecological Analogy for Generative Adversarial Networks and Diversity Control |
| title_sort | ecological analogy for generative adversarial networks and diversity control |
| topic | deep neural networks ecological dynamics generative adversarial networks training dynamics bio-inspired algorithm |
| url | https://doi.org/10.1088/2632-072X/acacdf |
| work_keys_str_mv | AT kenichinakazato ecologicalanalogyforgenerativeadversarialnetworksanddiversitycontrol |