Research map / Probabilistic and scientific modeling / Variational and generative sampling
Efficient Maximal Coding Rate Reduction by Variational Forms
Check 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 |
|---|---|---|---|
| replace the classwise log-determinant compression terms with the variational formulation | MCR² · MNIST | wall-clock time per epoch: 11.56 -> 6.29 | Helped |
| replace the classwise log-determinant compression terms with the variational formulation | MCR² · CIFAR-10 | wall-clock time per epoch: 33.06 -> 20.71 | Helped |
| replace the classwise log-determinant compression terms with the variational formulation | MCR² · CIFAR-100 | wall-clock time per epoch: 157.45 -> 31.14 | Helped |
| replace the classwise log-determinant compression terms with the variational formulation | MCR² · Tiny ImageNet | wall-clock time per epoch: 527.85 -> 44.23 | Helped |
| replace the original MCR² objective with V-MCR² | MCR² · MNIST | test accuracy: 0.9785 -> 0.9788 | No clear effect |
| replace the original MCR² objective with V-MCR² | MCR² · CIFAR-10 | test accuracy: 0.8956 -> 0.8997 | Helped |
| replace the original MCR² objective with V-MCR² | MCR² · CIFAR-100 | test accuracy: 0.2421 -> 0.5872 | Helped |
| replace the original MCR² objective with V-MCR² | MCR² · Tiny ImageNet | test accuracy: 0.1319 -> 0.2665 | Helped |
| use V-MCR² instead of CE under the additional training recipe | image classification training · CIFAR-100 | test accuracy: 0.7146 -> 0.6951 | Hurt |
| use V-MCR² instead of CE under the additional training recipe | image classification training · Tiny ImageNet | test accuracy: 0.4843 -> 0.4189 | Hurt |
Closest papers by meaning
- A Global Geometric Analysis of Maximal Coding Rate Reduction (ICML 2024 · similarity 0.89)
- Implicit Bias of Spectral Descent and Muon on Multiclass Separable Data (NeurIPS 2025 · similarity 0.84)
- Free Decompression with Algebraic Spectral Curves (NeurIPS 2026 · similarity 0.84)
- A Unified Framework for Learning with Nonlinear Model Classes from Arbitrary Linear Samples (ICML 2024 · similarity 0.84)
- An In-depth Investigation of Sparse Rate Reduction in Transformer-like Models (NeurIPS 2024 · similarity 0.83)
- Multilayer Matrix Factorization via Dimension-Reducing Diffusion Variational Inference (ICML 2025 · similarity 0.83)
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.