Research map / Time series / Causal and generative time series
A Study of Posterior Stability in Time-Series Latent Diffusion
Yangming Li, Yixin Cheng, Mihaela van der Schaar
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 latent diffusion with the proposed framework (diffusion as variational inference plus collapse simulation) | Latent diffusion, Transformer decoder · WARDS | Wasserstein distance: 7.46 -> 3.01 | Improved |
| replace latent diffusion with the proposed framework | Latent diffusion, LSTM decoder · MIMIC | Wasserstein distance: 5.19 -> 2.29 | Improved |
| add skip connections to latent diffusion | Latent diffusion, Transformer · WARDS | Wasserstein distance: 7.46 -> 4.67 | Improved |
| add KL annealing to latent diffusion | Latent diffusion, LSTM · Earthquakes | Wasserstein distance: 5.87 -> 3.88 | Improved |
| add variable masking to latent diffusion | Latent diffusion, Transformer · Earthquakes | Wasserstein distance: 5.91 -> 4.45 | Improved |
| increase collapse-penalty start M from 100 to 200 (ablation) | Proposed framework, Transformer · WARDS | Wasserstein distance: 3.01 -> 3.75 | Hurt |
| decrease collapse-penalty start M from 100 to 50 (ablation) | Proposed framework, Transformer · MIMIC | Wasserstein distance: 2.13 -> 2.59 | Hurt |
| increase variational-inference steps N from 50 to 150 (ablation) | Proposed framework, Transformer · MIMIC | Wasserstein distance: 2.13 -> 2.75 | Hurt |
| decrease N from 50 to 10 (ablation) | Proposed framework, Transformer · WARDS | Wasserstein distance: 3.01 -> 3.16 | Hurt |
| use proposed framework instead of latent diffusion on Retail/Energy | Latent diffusion, LSTM · Energy | MMD: 0.057 -> 0.033 | Improved |
| use proposed framework for text latent diffusion | Latent diffusion, Transformer · SNIPS | BLEU: 59.36 -> 78.12 | Improved |
| use proposed framework for image latent diffusion | Latent diffusion, U-Net · CIFAR-10 | FID: 3.91 -> 3.85 | No clear effect |
Closest papers by meaning
- Utilizing Image Transforms and Diffusion Models for Generative Modeling of Short and Long Time Series (NeurIPS 2024 · similarity 0.89)
- Smooth Diffusion: Crafting Smooth Latent Spaces in Diffusion Models (CVPR 2024 · similarity 0.89)
- Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders (ICLR 2024 · similarity 0.89)
- Masked Autoencoders Are Effective Tokenizers for Diffusion Models (ICML 2025 · similarity 0.89)
- DiffEnc: Variational Diffusion with a Learned Encoder (ICLR 2024 · similarity 0.89)
- Improving the Diffusability of Autoencoders (ICML 2025 · similarity 0.89)
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.
