Multiple Invertible and Partial-Equivariant Function for Latent Vector Transformation to Enhance Disentanglement in VAEs
Not on the map. Why?
Abstract (arXiv)
Disentanglement learning is central to understanding and reusing learned representations in variational autoencoders (VAEs). Although equivariance has been explored in this context, effectively exploiting it for disentanglement remains challenging. In this paper, we propose a novel method, called Multiple Invertible and Partial-Equivariant Transformation (MIPE-Transformation), which integrates two main parts: (1) Invertible and Partial-Equivariant Transformation (IPE-Transformation), guaranteeing an invertible latent-to-transformed-latent mapping while preserving partial input-to-latent equivariance in the transformed latent space; and (2) Exponential-Family Conversion (EF-Conversion) to extend the standard Gaussian prior to an approximate exponential family via a learnable conversion. In experiments on the 3D Cars, 3D Shapes, and dSprites datasets, MIPE-Transformation improves the disentanglement performance of state-of-the-art VAEs.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 31 of its 61 references that are in the map. First posted on arXiv 6 Feb 2025.
Builds on
31 of 61 references.
References · 31
- 2014Deep Learning Face Attributes in the Wildcited by 3,679
- 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
- 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 858
- 2018Disentangling by Factorisingcited by 614
- 2018Towards a Definition of Disentangled Representationscited by 294
- 2017Variational Inference of Disentangled Latent Concepts from Unlabeled Observationscited by 240
- 2020Weakly-Supervised Disentanglement Without Compromisescited by 237
- 2016Deep Unsupervised Clustering with Gaussian Mixture Variational Autoencoderscited by 219
- 2020Posterior Network: Uncertainty Estimation without OOD Samples via Density-Based Pseudo-Countscited by 163
- 2018Disentangled Sequential Autoencodercited by 117
- 2019Spatial Broadcast Decoder: A Simple Architecture for Learning Disentangled Representations in VAEscited by 116
- 2020Joint Disentangling and Adaptation for Cross-Domain Person Re-Identificationcited by 83
19 more
- 2020Understanding Failures of Deep Networks via Robust Feature Extractioncited by 56
- 2019Cross-Dataset Person Re-Identification via Unsupervised Pose Disentanglement and Adaptationcited by 44
- 2021Learning Partial Equivariances from Datacited by 44
- 2022Unsupervised Learning of Group Invariant and Equivariant Representationscited by 38
- 2021Learning Disentangled Representation by Exploiting Pretrained Generative Models: A Contrastive Learning Viewcited by 34
- 2021Contrastively Disentangled Sequential Variational Autoencodercited by 34
- 2020Progressive Learning and Disentanglement of Hierarchical Representationscited by 30
- 2021Commutative Lie Group VAE for Disentanglement Learningcited by 27
- 2022Symmetry-Based Representations for Artificial and Biological General Intelligencecited by 26
- 2019Learning Discrete and Continuous Factors of Data via Alternating Disentanglementcited by 23
- 2020Learning to Manipulate Individual Objects in an Imagecited by 22
- 2021Towards Building A Group-based Unsupervised Representation Disentanglement Frameworkcited by 22
- 2020The Convolution Exponential and Generalized Sylvester Flowscited by 18
- 2022Homomorphism Autoencoder -- Learning Group Structured Representations from Observed Transitionscited by 18
- 2014Deep Exponential Familiescited by 12
- 2021On Incorporating Inductive Biases into VAEscited by 11
- 2019Bayes-Factor-VAE: Hierarchical Bayesian Deep Auto-Encoder Models for Factor Disentanglementcited by 6
- 2020Generative Flows with Matrix Exponentialcited by 6
- 2024CFASL: Composite Factor-Aligned Symmetry Learning for Disentanglement in Variational AutoEncodercited by 2
Built on by
Not cited yet.