mlsanityData to 14 Oct 2026

Disentanglement in Difference: Directly Learning Semantically Disentangled Representations by Maximizing Inter-Factor Differences

arXiv 2502.03123 ↗First posted 5 Feb 2025cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

In this study, Disentanglement in Difference(DiD) is proposed to address the inherent inconsistency between the statistical independence of latent variables and the goal of semantic disentanglement in disentanglement representation learning. Conventional disentanglement methods achieve disentanglement representation by improving statistical independence among latent variables. However, the statistical independence of latent variables does not necessarily imply that they are semantically unrelated, thus, improving statistical independence does not always enhance disentanglement performance. To address the above issue, DiD is proposed to directly learn semantic differences rather than the statistical independence of latent variables. In the DiD, a Difference Encoder is designed to measure the semantic differences; a contrastive loss function is established to facilitate inter-dimensional comparison. Both of them allow the model to directly differentiate and disentangle distinct semantic factors, thereby resolving the inconsistency between statistical independence and semantic disentanglement. Experimental results on the dSprites and 3DShapes datasets demonstrate that the proposed DiD outperforms existing mainstream methods across various disentanglement metrics.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 18 of its 34 references that are in the map. First posted on arXiv 5 Feb 2025.

Builds on

18 of 34 references.

References · 18

  1. 2013Visualizing and Understanding Convolutional Networkscited by 2,290
  2. 2017Improved Training of Wasserstein GANscited by 2,235
  3. 2015Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networkscited by 2,054
  4. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 941
  5. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 860
  6. 2018Disentangling by Factorisingcited by 616
  7. 2018Towards a Definition of Disentangled Representationscited by 296
  8. 2017Variational Inference of Disentangled Latent Concepts from Unlabeled Observationscited by 241
  9. 2018Recent Advances in Autoencoder-Based Representation Learningcited by 133
  10. 2018Visual Object Networks: Image Generation with Disentangled 3D Representationcited by 124
  11. 2018Learning Deep Disentangled Embeddings with the F-Statistic Losscited by 101
  12. 2019Are Disentangled Representations Helpful for Abstract Visual Reasoning?cited by 100
6 more
  1. 2020Measuring Disentanglement: A Review of Metricscited by 70
  2. 2019Disentangling Factors of Variation Using Few Labelscited by 57
  3. 2019Demystifying Inter-Class Disentanglementcited by 31
  4. 2021Where and What? Examining Interpretable Disentangled Representationscited by 27
  5. 2020Evaluating the Disentanglement of Deep Generative Models through Manifold Topologycited by 20
  6. 2021Scaling-up Disentanglement for Image Translationcited by 15

Built on by

Not cited yet.