Research map / Point clouds / Point cloud classification
Turbo Autoencoder: Deep learning based channel codes for point-to-point communication channels
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 |
|---|---|---|---|
| Add interleaving to the TurboAE encoder (ablation) | TurboAE · AWGN, block length 100 | BER | Helped |
| Use interleaved encoding and iterative decoding instead of a plain CNN autoencoder | TurboAE · AWGN, moderate block length | BER | Helped |
| Use a continuous codeword rather than a binary codeword (ablation) | TurboAE · ATN channel, especially high SNR | BER | Helped |
| Use a continuous codeword rather than a binary codeword (ablation) | TurboAE · AWGN channel, high SNR | BER | Helped |
| Use a continuous codeword rather than a binary codeword (ablation) | TurboAE · Markovian-AWGN channel, high SNR | BER | Mixed |
| Pretrain with the continuous-power model before applying STE binarization (ablation) | TurboAE-binary · AWGN training | BER | Helped |
| Train encoder and decoder separately instead of simultaneously (ablation) | TurboAE · AWGN training | BER | Helped |
| Increase training batch size (ablation) | TurboAE · AWGN training | BER | Helped |
| Add residual connections to the decoder (ablation) | TurboAE · AWGN training | BER | Helped |
| Increase encoder and decoder network size (ablation) | TurboAE · AWGN training | BER | Helped |
| Fine-tune the model at block length 1000 rather than only testing a model trained at block length 100 (ablation) | TurboAE-continuous · AWGN, block length 1000 | BER | Helped |
| Train the encoder at an SNR matching the test SNR (ablation) | TurboAE · AWGN | BER | Helped |
Closest papers by meaning
- Communication Algorithms via Deep Learning (ICLR 2018 · similarity 0.88)
- Neural Joint Source-Channel Coding (ICML 2019 · similarity 0.87)
- Learning Linear Block Error Correction Codes (ICML 2024 · similarity 0.86)
- DeepPolar: Inventing Nonlinear Large-Kernel Polar Codes via Deep Learning (ICML 2024 · similarity 0.84)
- Friendly Attacks to Improve Channel Coding Reliability (AAAI 2024 · similarity 0.82)
- A Global Geometric Analysis of Maximal Coding Rate Reduction (ICML 2024 · similarity 0.81)
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.