Research map / Probabilistic and scientific modeling / Variational and generative sampling
A Generalised Linear Model Framework for β-Variational Autoencoders based on Exponential Dispersion Families
Robert Sicks, Ralf Korn, Stefanie Schwaar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| initialize with MLE-based weights (MLE-B) instead of standard rectifier init | beta-VAE, affine decoder · frey, synthetic, mnist (18 setups) | training convergence speed | Improved |
| initialize with MLE-based weights (MLE-B) instead of standard rectifier init | beta-VAE, affine decoder · frey, synthetic, mnist (18 setups) | final training objective | No clear effect |
| initialize with MLE-based weights (MLE-B) instead of standard rectifier init | beta-VAE, ReLU-Net decoder · frey, synthetic, mnist (18 setups) | training convergence speed | Mixed |
| replace affine decoder by ReLU-Net decoder | beta-VAE · MNIST Gaussian (beta=10) | histogram distance to analytical activity statistic: 0.05 -> 0.55 | Hurt |
| replace affine decoder by ReLU-Net decoder | beta-VAE · MNIST Gaussian (beta=1) | histogram distance to analytical activity statistic: 0.24 -> 0.18 | Hurt |
| increase beta from 1 to 2 | beta-VAE, affine decoder · MNIST Gaussian | histogram distance to analytical activity statistic: 0.24 -> 0.04 | Improved |
| increase beta from 10 to 20 | beta-VAE, ReLU-Net decoder · MNIST Gaussian | histogram distance to analytical activity statistic: 0.55 -> 0 | Improved |
| increase latent dimension to large values on synthetic data (ablation) | beta-VAE · synthetic | overfitting | No clear effect |
Closest papers by meaning
- Be a Goldfish: Forgetting Bad Conditioning in Sparse Linear Regression via Variational Autoencoders (ICML 2025 · similarity 0.90)
- Trading Information between Latents in Hierarchical Variational Autoencoders (ICLR 2023 · similarity 0.89)
- On Implicit Regularization in $β$-VAEs (ICML 2020 · similarity 0.89)
- Connections with Robust PCA and the Role of Emergent Sparsity in Variational Autoencoder Models (JMLR 2018 · similarity 0.89)
- Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders (ICLR 2024 · similarity 0.88)
- A Statistical Analysis of Wasserstein Autoencoders for Intrinsically Low-dimensional Data (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
