Learning disentangled representations with semi-supervised deep generative models
Variational autoencoders (VAEs) learn representations of data by jointly training a probabilistic encoder and decoder network. Typically these models encode all features of the data into a single variable. Here we are interested in learning disentangled representations that encode distinct aspects...
Main Authors: | , , , , , , , |
---|---|
Format: | Conference item |
Published: |
Curran Associates
2018
|
_version_ | 1797054505896902656 |
---|---|
author | Siddharth, N Paige, B Van De Meent, J Desmaison, A Goodman, N Kohli, P Wood, F Torr, P |
author_facet | Siddharth, N Paige, B Van De Meent, J Desmaison, A Goodman, N Kohli, P Wood, F Torr, P |
author_sort | Siddharth, N |
collection | OXFORD |
description | Variational autoencoders (VAEs) learn representations of data by jointly training a probabilistic encoder and decoder network. Typically these models encode all features of the data into a single variable. Here we are interested in learning disentangled representations that encode distinct aspects of the data into separate variables. We propose to learn such representations using model architectures that generalise from standard VAEs, employing a general graphical model structure in the encoder and decoder. This allows us to train partially-specified models that make relatively strong assumptions about a subset of interpretable variables and rely on the flexibility of neural networks to learn representations for the remaining variables. We further define a general objective for semi-supervised learning in this model class, which can be approximated using an importance sampling procedure. We evaluate our framework's ability to learn disentangled representations, both by qualitative exploration of its generative capacity, and quantitative evaluation of its discriminative ability on a variety of models and datasets. |
first_indexed | 2024-03-06T18:58:08Z |
format | Conference item |
id | oxford-uuid:128e9b3b-b303-4bdf-a849-72cab89b3635 |
institution | University of Oxford |
last_indexed | 2024-03-06T18:58:08Z |
publishDate | 2018 |
publisher | Curran Associates |
record_format | dspace |
spelling | oxford-uuid:128e9b3b-b303-4bdf-a849-72cab89b36352022-03-26T10:08:46ZLearning disentangled representations with semi-supervised deep generative modelsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:128e9b3b-b303-4bdf-a849-72cab89b3635Symplectic Elements at OxfordCurran Associates2018Siddharth, NPaige, BVan De Meent, JDesmaison, AGoodman, NKohli, PWood, FTorr, P Variational autoencoders (VAEs) learn representations of data by jointly training a probabilistic encoder and decoder network. Typically these models encode all features of the data into a single variable. Here we are interested in learning disentangled representations that encode distinct aspects of the data into separate variables. We propose to learn such representations using model architectures that generalise from standard VAEs, employing a general graphical model structure in the encoder and decoder. This allows us to train partially-specified models that make relatively strong assumptions about a subset of interpretable variables and rely on the flexibility of neural networks to learn representations for the remaining variables. We further define a general objective for semi-supervised learning in this model class, which can be approximated using an importance sampling procedure. We evaluate our framework's ability to learn disentangled representations, both by qualitative exploration of its generative capacity, and quantitative evaluation of its discriminative ability on a variety of models and datasets. |
spellingShingle | Siddharth, N Paige, B Van De Meent, J Desmaison, A Goodman, N Kohli, P Wood, F Torr, P Learning disentangled representations with semi-supervised deep generative models |
title | Learning disentangled representations with semi-supervised deep generative models |
title_full | Learning disentangled representations with semi-supervised deep generative models |
title_fullStr | Learning disentangled representations with semi-supervised deep generative models |
title_full_unstemmed | Learning disentangled representations with semi-supervised deep generative models |
title_short | Learning disentangled representations with semi-supervised deep generative models |
title_sort | learning disentangled representations with semi supervised deep generative models |
work_keys_str_mv | AT siddharthn learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT paigeb learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT vandemeentj learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT desmaisona learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT goodmann learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT kohlip learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT woodf learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels AT torrp learningdisentangledrepresentationswithsemisuperviseddeepgenerativemodels |