Research map / Label-efficient and robust learning / Recognition and long-tail
Generalised Mutual Information for Discriminative Clustering
Louis Ohl, Pierre-Alexandre Mattei, Charles Bouveyron, Warith HARCHAOUI, Mickaël Leclercq, Arnaud Droit, Frederic Precioso
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 KL mutual information with MMD OvA | Discriminative clustering on a Gaussian and Student-t mixture · Gaussian and Student-t mixture | ARI: 0 -> 0.922 | Improved |
| Use MMD OvO instead of MMD OvA (ablation) | GEMINI with an MLP · Gaussian and Student-t mixture | ARI: 0.922 -> 0.921 | No clear effect |
| Use Wasserstein OvO instead of Wasserstein OvA (ablation) | GEMINI with an MLP · Gaussian and Student-t mixture | ARI: 0.915 -> 0.922 | No clear effect |
| Replace KL mutual information with Wasserstein OvA | Discriminative clustering · MNIST, 10 clusters, MLP | ARI: 0.320 -> 0.471 | Improved |
| Replace KL mutual information with Wasserstein OvA | Discriminative clustering · MNIST, 10 clusters, LeNet-5 | ARI: 0.138 -> 0.463 | Improved |
| Replace KL mutual information with Wasserstein OvO | Discriminative clustering · MNIST, 15 clusters, MLP | ARI: 0.271 -> 0.415 | Improved |
| Replace KL mutual information with Wasserstein OvA | Discriminative clustering · MNIST, 15 clusters, LeNet-5 | ARI: 0.136 -> 0.446 | Improved |
| Change backbone from MLP to LeNet-5 (ablation) | KL OvA GEMINI · MNIST, 10 clusters | ARI: 0.320 -> 0.138 | Hurt |
| Replace linear kernel with a pretrained SIMCLR kernel (ablation) | MMD OvA GEMINI with LeNet-5 · CIFAR10 | ARI: 0.049 -> 0.157 | Improved |
| Replace Euclidean image distance with SIMCLR feature distance (ablation) | Wasserstein OvA GEMINI with LeNet-5 · CIFAR10 | ARI: 0.043 -> 0.079 | Improved |
| Replace linear kernel with a pretrained SIMCLR kernel (ablation) | MMD OvO GEMINI with ResNet-18 · CIFAR10 | ARI: 0.044 -> 0.145 | Improved |
| Replace Euclidean image distance with SIMCLR feature distance (ablation) | Wasserstein OvA GEMINI with ResNet-18 · CIFAR10 | ARI: 0.037 -> 0.052 | Improved |
Closest papers by meaning
- Mutual Information Estimation via $f$-Divergence and Data Derangements (NeurIPS 2024 · similarity 0.85)
- The Map Equation Goes Neural: Mapping Network Flows with Graph Neural Networks (NeurIPS 2024 · similarity 0.85)
- Slicing Mutual Information Generalization Bounds for Neural Networks (ICML 2024 · similarity 0.85)
- Connecting Jensen–Shannon and Kullback–Leibler Divergences: A New Bound for Representation Learning (NeurIPS 2025 · similarity 0.85)
- Scalable Multiple Kernel Clustering: Learning Clustering Structure from Expectation (ICML 2024 · similarity 0.85)
- InfoNet: Neural Estimation of Mutual Information without Test-Time Optimization (ICML 2024 · 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.
