The paper proposes a PCA-based method to find latent directions for each generative factor, creating disentanglement metrics that work for latent variable models with non-axis-aligned encodings, and reports improved scores on DSprites and 3D Shapes.
Recent advances in autoencoder-based representation learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Disentanglement Analysis in Deep Latent Variable Models Matching Aggregate Posterior Distributions
The paper proposes a PCA-based method to find latent directions for each generative factor, creating disentanglement metrics that work for latent variable models with non-axis-aligned encodings, and reports improved scores on DSprites and 3D Shapes.