It takes (only) two: adversarial generator-encoder networks
We present a new autoencoder-type architecture that is trainable in an unsupervised mode, sustains both generation and inference, and has the quality of conditional and unconditional samples boosted by adversarial learning. Unlike previous hybrids of autoencoders and adversarial networks, the advers...
| Príomhchruthaitheoirí: | , , |
|---|---|
| Formáid: | Conference item |
| Foilsithe / Cruthaithe: |
Association for the Advancement of Artificial Intelligence
2018
|
| _version_ | 1826268082217156608 |
|---|---|
| author | Ulyanov, D Vedaldi, A Lempitsky, V |
| author_facet | Ulyanov, D Vedaldi, A Lempitsky, V |
| author_sort | Ulyanov, D |
| collection | OXFORD |
| description | We present a new autoencoder-type architecture that is trainable in an unsupervised mode, sustains both generation and inference, and has the quality of conditional and unconditional samples boosted by adversarial learning. Unlike previous hybrids of autoencoders and adversarial networks, the adversarial game in our approach is set up directly between the encoder and the generator, and no external mappings are trained in the process of learning. The game objective compares the divergences of each of the real and the generated data distributions with the prior distribution in the latent space. We show that direct generator-vs-encoder game leads to a tight coupling of the two components, resulting in samples and reconstructions of a comparable quality to some recently-proposed more complex architectures. |
| first_indexed | 2024-03-06T21:04:08Z |
| format | Conference item |
| id | oxford-uuid:3be3099a-b8c5-4c61-a39c-c5c950a42a6d |
| institution | University of Oxford |
| last_indexed | 2024-03-06T21:04:08Z |
| publishDate | 2018 |
| publisher | Association for the Advancement of Artificial Intelligence |
| record_format | dspace |
| spelling | oxford-uuid:3be3099a-b8c5-4c61-a39c-c5c950a42a6d2022-03-26T14:10:09ZIt takes (only) two: adversarial generator-encoder networksConference itemhttp://purl.org/coar/resource_type/c_5794uuid:3be3099a-b8c5-4c61-a39c-c5c950a42a6dSymplectic Elements at OxfordAssociation for the Advancement of Artificial Intelligence2018Ulyanov, DVedaldi, ALempitsky, VWe present a new autoencoder-type architecture that is trainable in an unsupervised mode, sustains both generation and inference, and has the quality of conditional and unconditional samples boosted by adversarial learning. Unlike previous hybrids of autoencoders and adversarial networks, the adversarial game in our approach is set up directly between the encoder and the generator, and no external mappings are trained in the process of learning. The game objective compares the divergences of each of the real and the generated data distributions with the prior distribution in the latent space. We show that direct generator-vs-encoder game leads to a tight coupling of the two components, resulting in samples and reconstructions of a comparable quality to some recently-proposed more complex architectures. |
| spellingShingle | Ulyanov, D Vedaldi, A Lempitsky, V It takes (only) two: adversarial generator-encoder networks |
| title | It takes (only) two: adversarial generator-encoder networks |
| title_full | It takes (only) two: adversarial generator-encoder networks |
| title_fullStr | It takes (only) two: adversarial generator-encoder networks |
| title_full_unstemmed | It takes (only) two: adversarial generator-encoder networks |
| title_short | It takes (only) two: adversarial generator-encoder networks |
| title_sort | it takes only two adversarial generator encoder networks |
| work_keys_str_mv | AT ulyanovd ittakesonlytwoadversarialgeneratorencodernetworks AT vedaldia ittakesonlytwoadversarialgeneratorencodernetworks AT lempitskyv ittakesonlytwoadversarialgeneratorencodernetworks |