Research map / Probabilistic and scientific modeling / Variational and generative sampling
A New Distribution on the Simplex with Auto-Encoding Applications
Andrew Stirn, Tony Jebara, David Knowles
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 |
|---|---|---|---|
| Replace fixed-order Kumaraswamy stick-breaking with averaging over sampled orderings | Kumar-SB semi-supervised model · MNIST, 600 labels, dim(z)=50 | classification error: 0.071 ± 0.008 -> 0.018 ± 0.004 | Improved |
| Replace fixed-order Kumaraswamy stick-breaking with averaging over sampled orderings | Kumar-SB semi-supervised model · SVHN, 1000 labels, dim(z)=50 | classification error: 0.702 ± 0.011 -> 0.296 ± 0.014 | Improved |
| Use MV-Kum. instead of IRG | Semi-supervised model · MNIST, 600 labels, dim(z)=50 | classification error: 0.018 ± 0.004 vs 0.018 ± 0.004 | No clear effect |
| Use MV-Kum. instead of IRG | Semi-supervised model · SVHN, 1000 labels, dim(z)=50 | classification error: 0.296 ± 0.014 vs 0.288 ± 0.008 | No clear effect |
| Use MV-Kum. instead of Softmax | Semi-supervised model · MNIST, 600 labels, dim(z)=2 | classification error: 0.043 ± 0.005 vs 0.042 ± 0.003 | No clear effect |
| Use MV-Kum. instead of Softmax | Semi-supervised model · SVHN, 1000 labels, dim(z)=50 | classification error: 0.296 ± 0.014 vs 0.300 ± 0.007 | No clear effect |
| Use MV-Kum. rather than the stacked M1 + M2 model | Semi-supervised model · SVHN, 1000 labels | classification error: 0.360 ± 0.001 -> 0.296 ± 0.014 | Improved |
| Use dim(z)=50 instead of dim(z)=2 | MV-Kum. on MNIST · MNIST, 600 labels | classification error: 0.043 ± 0.005 -> 0.018 ± 0.004 | Improved |
| Use dim(z)=0 instead of dim(z)=50 | MV-Kum. on MNIST · MNIST, 600 labels | classification error: 0.018 ± 0.004 -> 0.099 ± 0.011 | Hurt |
Closest papers by meaning
- The continuous categorical: a novel simplex-valued exponential family (ICML 2020 · similarity 0.87)
- Stick-Breaking Variational Autoencoders (ICLR 2017 · similarity 0.86)
- The continuous Bernoulli: fixing a pervasive error in variational autoencoders (NeurIPS 2019 · similarity 0.86)
- Adversarial Symmetric Variational Autoencoder (NeurIPS 2017 · similarity 0.85)
- Generalization in VAE and Diffusion Models: A Unified Information-Theoretic Analysis (ICLR 2025 · similarity 0.84)
- $t^3$-Variational Autoencoder: Learning Heavy-tailed Data with Student's t and Power Divergence (ICLR 2024 · similarity 0.84)
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.
