Research map / Probabilistic and scientific modeling / Variational and generative sampling
Evolutionary Variational Optimization of Generative Models
Jakob Drefs, Enrico Guiraud, Jörg Lücke
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 |
|---|---|---|---|
| use spike-and-slab SSSC model instead of BSC under EEM | EEM denoising · House σ=15 | PSNR (dB): 33.66 -> 34.90 | Improved |
| use SSSC instead of BSC under EEM | EEM denoising · House σ=25 | PSNR (dB): 32.40 -> 33.15 | Improved |
| use SSSC instead of BSC under EEM | EEM denoising · House σ=50 | PSNR (dB): 28.98 -> 29.83 | Improved |
| use SSSC instead of BSC under EEM | EEM denoising · Barbara σ=25 | PSNR (dB): 28.38 -> 30.19 | Improved |
| use SSSC instead of BSC under EEM | EEM denoising · Lena σ=25 | PSNR (dB): 30.88 -> 31.95 | Improved |
| use SSSC instead of BSC under EEM | EEM denoising · Peppers σ=25 | PSNR (dB): 28.93 -> 30.27 | Improved |
| different genetic operator combinations for the EA (ablation) | EEM on artificial bars data · bars test ground-truth recovery | PSNR (dB) | Mixed |
| train SSSC with EEM instead of mean-field (MTMKL), sampling (BPFA) or preselection-only truncation (GSC) | SSSC denoising · House σ=50, controlled D and H | PSNR (dB) | Improved |
| use EEM-trained BSC (EBSC) instead of MTMKL | BSC denoising · House σ=50, controlled D and H | PSNR (dB) | Improved |
| run the variational loop longer (higher lower bound) during training | ES3C · House σ=50 PSNR vs variational bound | PSNR (dB): 28.94, 28.95, 28.94, 28.96, 28.95 | Mixed |
Closest papers by meaning
- Variational Diffusion Models (NeurIPS 2021 · similarity 0.89)
- From Variational to Deterministic Autoencoders (ICLR 2020 · similarity 0.89)
- Variational Search Distributions (ICLR 2025 · similarity 0.87)
- Aligning Few-Step Generative Model via Amortizing Sample-Based Variational Inference (NeurIPS 2026 · similarity 0.87)
- Posterior Inference in Latent Space for Scalable Constrained Black-box Optimization (NeurIPS 2026 · similarity 0.87)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.87)
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.
