Dropout inference in Bayesian neural networks with alpha-divergences
To obtain uncertainty estimates with real-world Bayesian deep learning models, practical inference approximations are needed. Dropout variational inference (VI) for example has been used for machine vision and medical applications, but VI can severely underestimates model uncertainty. Alpha-divergen...
| Những tác giả chính: | , |
|---|---|
| Định dạng: | Conference item |
| Được phát hành: |
PMLR
2017
|
| _version_ | 1826304044049629184 |
|---|---|
| author | Li, Y Gal, Y |
| author_facet | Li, Y Gal, Y |
| author_sort | Li, Y |
| collection | OXFORD |
| description | To obtain uncertainty estimates with real-world Bayesian deep learning models, practical inference approximations are needed. Dropout variational inference (VI) for example has been used for machine vision and medical applications, but VI can severely underestimates model uncertainty. Alpha-divergences are alternative divergences to VI’s KL objective, which are able to avoid VI’s uncertainty underestimation. But these are hard to use in practice: existing techniques can only use Gaussian approximating distributions, and require existing models to be changed radically, thus are of limited use for practitioners. We propose a re-parametrisation of the alpha-divergence objectives, deriving a simple inference technique which, together with dropout, can be easily implemented with existing models by simply changing the loss of the model. We demonstrate improved uncertainty estimates and accuracy compared to VI in dropout networks. We study our model’s epistemic uncertainty far away from the data using adversarial images, showing that these can be distinguished from non-adversarial images by examining our model’s uncertainty. |
| first_indexed | 2024-03-07T06:11:51Z |
| format | Conference item |
| id | oxford-uuid:efc803b6-6b67-4481-8e4d-d945b44ebd9d |
| institution | University of Oxford |
| last_indexed | 2024-03-07T06:11:51Z |
| publishDate | 2017 |
| publisher | PMLR |
| record_format | dspace |
| spelling | oxford-uuid:efc803b6-6b67-4481-8e4d-d945b44ebd9d2022-03-27T11:42:50ZDropout inference in Bayesian neural networks with alpha-divergencesConference itemhttp://purl.org/coar/resource_type/c_5794uuid:efc803b6-6b67-4481-8e4d-d945b44ebd9dSymplectic Elements at OxfordPMLR2017Li, YGal, YTo obtain uncertainty estimates with real-world Bayesian deep learning models, practical inference approximations are needed. Dropout variational inference (VI) for example has been used for machine vision and medical applications, but VI can severely underestimates model uncertainty. Alpha-divergences are alternative divergences to VI’s KL objective, which are able to avoid VI’s uncertainty underestimation. But these are hard to use in practice: existing techniques can only use Gaussian approximating distributions, and require existing models to be changed radically, thus are of limited use for practitioners. We propose a re-parametrisation of the alpha-divergence objectives, deriving a simple inference technique which, together with dropout, can be easily implemented with existing models by simply changing the loss of the model. We demonstrate improved uncertainty estimates and accuracy compared to VI in dropout networks. We study our model’s epistemic uncertainty far away from the data using adversarial images, showing that these can be distinguished from non-adversarial images by examining our model’s uncertainty. |
| spellingShingle | Li, Y Gal, Y Dropout inference in Bayesian neural networks with alpha-divergences |
| title | Dropout inference in Bayesian neural networks with alpha-divergences |
| title_full | Dropout inference in Bayesian neural networks with alpha-divergences |
| title_fullStr | Dropout inference in Bayesian neural networks with alpha-divergences |
| title_full_unstemmed | Dropout inference in Bayesian neural networks with alpha-divergences |
| title_short | Dropout inference in Bayesian neural networks with alpha-divergences |
| title_sort | dropout inference in bayesian neural networks with alpha divergences |
| work_keys_str_mv | AT liy dropoutinferenceinbayesianneuralnetworkswithalphadivergences AT galy dropoutinferenceinbayesianneuralnetworkswithalphadivergences |