Research map / Probabilistic and scientific modeling / Variational and generative sampling
Neural Joint Source-Channel Coding
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 |
|---|---|---|---|
| Train NECST jointly for source and channel coding instead of separating compression and error correction | Image coding · CIFAR-10, CelebA, binarized SVHN | bits required at matched distortion: 20× less bits | Helped |
| Train NECST jointly rather than use discrete VAE plus rate-1/2 LDPC | Image reconstruction · Binarized MNIST and Omniglot | distortion | Helped |
| Train features with simulated channel noise | NECST representations · MNIST and noisy MNIST classification | accuracy | Helped |
| Train features with simulated channel noise | NECST representations · MNIST classification with KNN | accuracy: 0.95 -> 0.95; 0.95 -> 0.94 | No clear effect |
| Replace discrete latent features with a continuous relaxation (ablation) | NECST training · Test-time discrete codes | performance | Hurt |
| Use a learned amortized neural decoder instead of iterative belief-propagation decoding | NECST vs. LDPC · Omniglot | decoding speed | Helped |
Closest papers by meaning
- Infomax Neural Joint Source-Channel Coding via Adversarial Bit Flip (AAAI 2020 · similarity 0.89)
- Learning Linear Block Error Correction Codes (ICML 2024 · similarity 0.86)
- Learned Image Transmission with Hierarchical Variational Autoencoder (AAAI 2025 · similarity 0.85)
- Approaching Rate-Distortion Limits in Neural Compression with Lattice Transform Coding (ICLR 2025 · similarity 0.84)
- Structured Masked Diffusion for Joint Multiuser Decoding (NeurIPS 2026 · similarity 0.84)
- Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders (ICML 2025 · 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.