Research map / Label-efficient and robust learning / Recognition and long-tail
Scalable Infomin Learning
Yanzhi Chen, Weihao Sun, Yingzhen Li, Adrian Weller
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 adversarial MI estimators with sliced (Slice) estimator | infomin representation learning · US Census Demographic | fairness ρ*(Z,T): 0.16 \pm 0.10 -> 0.07 \pm 0.02 | Improved |
| use Slice instead of Neural Rényi | infomin representation learning · UCI Adult | fairness ρ*(Z,T): 0.17 \pm 0.08 -> 0.08 \pm 0.02 | Improved |
| use Slice instead of dCorr/Pearson non-parametric penalties | infomin representation learning · US Census Demographic | fairness ρ*(Z,T): 0.47 \pm 0.08 -> 0.07 \pm 0.02 | Improved |
| apply Slice infomin term | infomin representation learning · Dsprite | ρ*(Z,T): 0.91 \pm 0.03 -> 0.08 \pm 0.02 | Improved |
| apply Slice infomin term | infomin representation learning · Dsprite | reconstruction MSE: 0.37 \pm 0.01 -> 0.50 \pm 0.01 | Hurt |
| apply Slice infomin term | infomin representation learning · CMU-PIE | ρ*(Z,T): 0.76 \pm 0.04 -> 0.07 \pm 0.01 | Improved |
| apply Slice infomin term | infomin representation learning · CMU-PIE | reconstruction MSE: 1.81 \pm 0.04 -> 2.15 \pm 0.07 | Hurt |
| use Slice instead of adversarial/Neural TC | infomin representation learning · MNIST->MNIST-M | target accuracy: 46.3 \pm 0.03 -> 98.5 \pm 0.45 | Improved |
| use Slice instead of Neural TC | infomin representation learning · MNIST->MNIST-M | ρ*(Zc,Zd): 0.64 \pm 0.04 -> 0.06 \pm 0.01 | Improved |
| use Slice instead of Neural TC | infomin representation learning · CIFAR10->STL10 | target accuracy: 80.8 \pm 0.08 -> 82.3 \pm 0.03 | Improved |
| apply Slice infomin term | infomin representation learning · CIFAR10->STL10 | source accuracy: 93.0 \pm 0.03 -> 92.5 \pm 0.03 | Hurt |
| apply Slice infomin term | infomin representation learning · US Census Demographic | utility ρ*(Z,Y): 0.95 \pm 0.00 -> 0.95 \pm 0.01 | No clear effect |
Closest papers by meaning
- InfoNet: Neural Estimation of Mutual Information without Test-Time Optimization (ICML 2024 · similarity 0.85)
- On Mutual Information Maximization for Representation Learning (ICLR 2020 · similarity 0.85)
- Inherent Tradeoffs in Learning Fair Representations (NeurIPS 2019 · similarity 0.84)
- Mutual Information Neural Estimation (ICML 2018 · similarity 0.84)
- Impartial Adversarial Distillation: Addressing Biased Data-Free Knowledge Distillation via Adaptive Constrained Optimization (AAAI 2024 · similarity 0.83)
- Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference (NeurIPS 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.
