Research map / Probabilistic and scientific modeling / Variational and generative sampling
Continuous Mixtures of Tractable Probabilistic Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use continuous latent mixtures of Chow-Liu trees instead of factorised components (ablation) | Continuous mixtures of tractable models · 20 density-estimation benchmarks | test log-likelihood | Helped |
| Train the decoder by numerical integration rather than variational inference (ablation) | Continuous mixtures with factorised structure · 20 density-estimation benchmarks | test log-likelihood | Helped |
| Optimize integration points after decoder training (ablation) | Continuous mixtures of Chow-Liu trees · 20 density-estimation benchmarks | test log-likelihood | No clear effect |
| Use latent optimization with too many trainable integration points relative to training data (ablation) | Continuous mixtures · Density-estimation benchmarks | test log-likelihood | Hurt |
| Use normal rather than categorical leaf distributions (ablation) | Continuous mixtures of factorised models · MNIST, Fashion-MNIST, SVHN | bits per dimension | Hurt |
| Use RQMC rather than MC integration during training (ablation) | Continuous mixtures of factorised models · Binary MNIST | validation log-likelihood | Helped |
| Increase latent dimension from 1 to 4 (ablation) | cmf · Binary MNIST | test log-likelihood: -128.05 -> -106.97 | Helped |
| Increase latent dimension from 4 to 8 (ablation) | cmf · Binary MNIST | test log-likelihood: -106.97 -> -103.82 | Helped |
| Increase latent dimension from 8 to 64 (ablation) | cmf · Binary MNIST | test log-likelihood: -103.82 -> -104.21 | No clear effect |
| Increase latent dimension from 4 to 8 (ablation) | cmvae · Binary MNIST | test log-likelihood: -111.57 -> -110.67 | Helped |
| Increase latent dimension from 8 to 64 (ablation) | cmvae · Binary MNIST | test log-likelihood: -110.67 -> -116.57 | Hurt |
| Replace shared decoder-generated mixture components with independent plain-mixture components (ablation) | Factorised mixture model · 20 density-estimation benchmarks | test log-likelihood | Hurt |
Closest papers by meaning
- Efficient Mixture Learning in Black-Box Variational Inference (ICML 2024 · similarity 0.87)
- Scaling Continuous Latent Variable Models as Probabilistic Integral Circuits (NeurIPS 2024 · similarity 0.86)
- Sum of Squares Circuits (AAAI 2025 · similarity 0.86)
- On the Hardness of Approximating Distributions with Tractable Probabilistic Models (NeurIPS 2025 · similarity 0.86)
- Variational Learning of Fractional Posteriors (ICML 2025 · similarity 0.86)
- Subtractive Mixture Models via Squaring: Representation and Learning (ICLR 2024 · similarity 0.86)
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.