Research map / Architectures and efficiency / RNNs and normalization
Lossless Compression with Probabilistic Circuits
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 |
|---|---|---|---|
| Use an HCLT probabilistic circuit for direct lossless compression | MNIST · MNIST test | codeword bpd: 1.31 -> 1.24 | Helped |
| Use an HCLT probabilistic circuit instead of BitSwap | FashionMNIST · FashionMNIST | codeword bpd: 3.35 -> 3.37 | Hurt |
| Use an HCLT probabilistic circuit for direct lossless compression | EMNIST · EMNIST (Letter) | codeword bpd: 2.02 -> 1.84 | Helped |
| Use an HCLT probabilistic circuit for direct lossless compression | EMNIST · EMNIST (ByClass) | codeword bpd: 2.04 -> 1.89 | Helped |
| Replace IDF's independent latent distributions with probabilistic circuits | IDF · CIFAR10 | theoretical bpd: 3.32 -> 3.28 | Helped |
| Replace IDF's independent latent distributions with probabilistic circuits | IDF · ImageNet32 | theoretical bpd: 4.15 -> 3.99 | Helped |
| Replace IDF's independent latent distributions with probabilistic circuits | IDF · ImageNet64 | theoretical bpd: 3.90 -> 3.71 | Helped |
| Increase HCLT size from M=16 to M=32 (ablation) | PC (HCLT) · MNIST test | codeword bpd: 1.30 -> 1.24 | Helped |
| Increase HCLT size from M=16 to M=32 (ablation) | PC (HCLT) · MNIST test | compression time: 9 -> 26 | Hurt |
| Use structured-decomposable PC marginal computation rather than the naive approach (ablation) | PC lossless compressor · D-dimensional data | computation time complexity: O(D · |p|) -> O(log(D) · |p|) | Helped |
Closest papers by meaning
- Practical Full Resolution Learned Lossless Image Compression (CVPR 2019 · similarity 0.89)
- Sparse Probabilistic Circuits via Pruning and Growing (NeurIPS 2022 · similarity 0.88)
- Lossy Compression for Lossless Prediction (NeurIPS 2021 · similarity 0.87)
- Learning Better Lossless Compression Using Lossy Compression (CVPR 2020 · similarity 0.87)
- Practical lossless compression with latent variables using bits back coding (ICLR 2019 · similarity 0.87)
- Robustly overfitting latents for flexible neural image compression (NeurIPS 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.