Research map / Learning theory and algorithms / Prediction and decision losses
Data-Efficient Learning via Minimizing Hyperspherical Energy
Xiaofeng Cao, Weiyang Liu, Ivor Wai-Hung Tsang
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 |
|---|---|---|---|
| use SphericalKmeans instead of K-means for deep clustering (ablation) | K-means+Auto+DEC · MNIST clustering | optimal accuracy: 0.96221 -> 0.96864 | Improved |
| use SphericalKmeans instead of GMM (ablation) | GMM+Auto+DEC · Fashion-MNIST clustering | optimal accuracy: 0.61319 -> 0.63448 | Improved |
| add AutoEncoder and DEC loss to spherical K-means (ablation) | SphericalKmeans · MNIST clustering | optimal accuracy: 0.55188 -> 0.96864 | Improved |
| use l0 MHE instead of l2 MHE (ablation) | l2 MHE · MNIST distribution matching | KL mean loss: 0.2751 -> 0.2183 | Improved |
| use l0 MHE instead of l1 MHE (ablation) | l1 MHE · MNIST distribution matching | KL mean loss: 0.2103 -> 0.2183 | Hurt |
| use l0 MHE instead of l2 MHE (ablation) | l2 MHE · Fashion-MNIST distribution matching | MMD_mu mean loss: 0.0410 -> 0.0364 | Improved |
| sample representative data in in-version-space instead of out-of-version-space (ablation) | random sampling · MNIST and Fashion-MNIST | classification accuracy | Improved |
| MHEAL versus Core-set in deep AL | Core-set · MNIST 100 labels | accuracy: 0.6310 -> 0.8648 | Improved |
| MHEAL versus Core-set in deep AL | Core-set · CIFAR-100 2K labels | accuracy: 0.2645 -> 0.3201 | Improved |
| MHEAL versus BALD at 2K CIFAR-10 labels | BALD · CIFAR-10 2K labels | accuracy: 0.6203 -> 0.6510 | Hurt |
| add 10,000 repeated samples to training set | MHEAL · CIFAR-10 10K labels | accuracy: 0.8146 -> 0.8042 | Hurt |
| add 10,000 noisy samples to training set | MHEAL · CIFAR-10 10K labels | accuracy: 0.8146 -> 0.7885 | Hurt |
Closest papers by meaning
- Learning towards Minimum Hyperspherical Energy (NeurIPS 2018 · similarity 0.85)
- SPHERE-JEPA: Spherical Prediction with Homogeneous Embeddings (NeurIPS 2026 · similarity 0.85)
- Fast Hyperboloid Decision Tree Algorithms (ICLR 2024 · similarity 0.84)
- Riemannian Geometric-based Meta Learning (AAAI 2025 · similarity 0.84)
- Hardness of Learning Neural Networks under the Manifold Hypothesis (NeurIPS 2024 · similarity 0.83)
- Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves (NeurIPS 2026 · 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.
