Research map / Probabilistic and scientific modeling / Variational and generative sampling
Convolutional dictionary learning based auto-encoders for natural exponential-family distributions
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 |
|---|---|---|---|
| untie encoder and decoder weights, allowing more parameters (ablation) | DCEA-C · Poisson image denoising, Set12 and BSD68 | PSNR (dB): DCEA-C to DCEA-UC: Set12 peak 1 20.72 -> 21.37; BSD68 peak 1 21.27 -> 21.84; Set12 peak 2 22.02 -> 22.79; BSD68 peak 2 2 | Helped |
| replace the linear decoder with a nonlinear two-layer decoder (ablation) | DCEA-UC · Poisson image denoising | PSNR (dB): increase in PSNR of ~0.2 dB | Helped |
| untie encoder and decoder weights (ablation) | DCEA-C · Simulated binomial ECDL | filter recovery error | Hurt |
| use a linear link and Gaussian likelihood instead of the correct sigmoid link and binomial likelihood (ablation) | DCEA-b · Simulated binomial ECDL, 200 observations | filter recovery error: DCEA-g with 200 observations: close to 0.4; DCEA-b with M_j=30: 0.09 | Hurt |
| use the correct binomial likelihood and sigmoid link (ablation) | DCEA-g · Simulated binomial ECDL | filter recovery error: close to 0.4 -> 0.09 | Helped |
| replace the ReLU sparse-code operator with hard thresholding (ablation) | Binomial SEA · Binomial dictionary recovery under the theorem assumptions | dictionary recovery | No clear effect |
Closest papers by meaning
- On the Sparsity-Storage-Accuracy Tradeoff in Parsimoniously Activated Dictionary Learning (NeurIPS 2026 · similarity 0.86)
- Denoising Autoregressive Representation Learning (ICML 2024 · similarity 0.86)
- Unified Generation, Reconstruction, and Representation: Generalized Diffusion with Adaptive Latent Encoding-Decoding (ICML 2024 · similarity 0.85)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.85)
- End-to-End Training for Unified Tokenization and Latent Denoising (NeurIPS 2026 · similarity 0.85)
- Sparse Autoencoders, Again? (ICML 2025 · similarity 0.85)
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.