Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387432640> ?p ?o ?g. }
Showing items 1 to 67 of
67
with 100 items per page.
- W4387432640 abstract "Variational dimensionality reduction methods are known for their high accuracy, generative abilities, and robustness. These methods have many theoretical justifications. Here we introduce a unifying principle rooted in information theory to rederive and generalize existing variational methods and design new ones. We base our framework on an interpretation of the multivariate information bottleneck, in which two Bayesian networks are traded off against one another. We interpret the first network as an encoder graph, which specifies what information to keep when compressing the data. We interpret the second network as a decoder graph, which specifies a generative model for the data. Using this framework, we rederive existing dimensionality reduction methods such as the deep variational information bottleneck (DVIB), beta variational auto-encoders (beta-VAE), and deep variational canonical correlation analysis (DVCCA). The framework naturally introduces a trade-off parameter between compression and reconstruction in the DVCCA family of algorithms, resulting in the new beta-DVCCA family. In addition, we derive a new variational dimensionality reduction method, deep variational symmetric informational bottleneck (DVSIB), which simultaneously compresses two variables to preserve information between their compressed representations. We implement all of these algorithms and evaluate their ability to produce shared low dimensional latent spaces on a modified noisy MNIST dataset. We show that algorithms that are better matched to the structure of the data (beta-DVCCA and DVSIB) produce better latent spaces as measured by classification accuracy and the dimensionality of the latent variables. We believe that this framework can be used to unify other multi-view representation learning algorithms. Additionally, it provides a straightforward framework for deriving problem-specific loss functions." @default.
- W4387432640 created "2023-10-09" @default.
- W4387432640 creator A5030584398 @default.
- W4387432640 creator A5071388278 @default.
- W4387432640 creator A5093023319 @default.
- W4387432640 date "2023-10-05" @default.
- W4387432640 modified "2023-10-09" @default.
- W4387432640 title "Deep Variational Multivariate Information Bottleneck -- A Framework for Variational Losses" @default.
- W4387432640 doi "https://doi.org/10.48550/arxiv.2310.03311" @default.
- W4387432640 hasPublicationYear "2023" @default.
- W4387432640 type Work @default.
- W4387432640 citedByCount "0" @default.
- W4387432640 crossrefType "posted-content" @default.
- W4387432640 hasAuthorship W4387432640A5030584398 @default.
- W4387432640 hasAuthorship W4387432640A5071388278 @default.
- W4387432640 hasAuthorship W4387432640A5093023319 @default.
- W4387432640 hasBestOaLocation W43874326401 @default.
- W4387432640 hasConcept C108583219 @default.
- W4387432640 hasConcept C111030470 @default.
- W4387432640 hasConcept C11413529 @default.
- W4387432640 hasConcept C119857082 @default.
- W4387432640 hasConcept C126255220 @default.
- W4387432640 hasConcept C149635348 @default.
- W4387432640 hasConcept C152139883 @default.
- W4387432640 hasConcept C154945302 @default.
- W4387432640 hasConcept C167966045 @default.
- W4387432640 hasConcept C190502265 @default.
- W4387432640 hasConcept C2780513914 @default.
- W4387432640 hasConcept C33923547 @default.
- W4387432640 hasConcept C39890363 @default.
- W4387432640 hasConcept C41008148 @default.
- W4387432640 hasConcept C60008888 @default.
- W4387432640 hasConcept C70518039 @default.
- W4387432640 hasConcept C80444323 @default.
- W4387432640 hasConceptScore W4387432640C108583219 @default.
- W4387432640 hasConceptScore W4387432640C111030470 @default.
- W4387432640 hasConceptScore W4387432640C11413529 @default.
- W4387432640 hasConceptScore W4387432640C119857082 @default.
- W4387432640 hasConceptScore W4387432640C126255220 @default.
- W4387432640 hasConceptScore W4387432640C149635348 @default.
- W4387432640 hasConceptScore W4387432640C152139883 @default.
- W4387432640 hasConceptScore W4387432640C154945302 @default.
- W4387432640 hasConceptScore W4387432640C167966045 @default.
- W4387432640 hasConceptScore W4387432640C190502265 @default.
- W4387432640 hasConceptScore W4387432640C2780513914 @default.
- W4387432640 hasConceptScore W4387432640C33923547 @default.
- W4387432640 hasConceptScore W4387432640C39890363 @default.
- W4387432640 hasConceptScore W4387432640C41008148 @default.
- W4387432640 hasConceptScore W4387432640C60008888 @default.
- W4387432640 hasConceptScore W4387432640C70518039 @default.
- W4387432640 hasConceptScore W4387432640C80444323 @default.
- W4387432640 hasLocation W43874326401 @default.
- W4387432640 hasOpenAccess W4387432640 @default.
- W4387432640 hasPrimaryLocation W43874326401 @default.
- W4387432640 hasRelatedWork W2002563186 @default.
- W4387432640 hasRelatedWork W2906272760 @default.
- W4387432640 hasRelatedWork W2978098801 @default.
- W4387432640 hasRelatedWork W2980541498 @default.
- W4387432640 hasRelatedWork W3014948380 @default.
- W4387432640 hasRelatedWork W3137091086 @default.
- W4387432640 hasRelatedWork W4289105138 @default.
- W4387432640 hasRelatedWork W4310699748 @default.
- W4387432640 hasRelatedWork W4313338966 @default.
- W4387432640 hasRelatedWork W4365211920 @default.
- W4387432640 isParatext "false" @default.
- W4387432640 isRetracted "false" @default.
- W4387432640 workType "article" @default.