Research map / Label-efficient and robust learning / Recognition and long-tail
Invariant-Equivariant Representation Learning for Multi-Class Data
Ilya Feige
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 the invariant embedding for classification instead of the dedicated benchmark classifier | EquiVAE on MNIST · MNIST test (supervised) | error rate: 0.84 ± 0.03 -> 0.82 ± 0.03 | Improved |
| Use the invariant embedding for classification instead of the dedicated benchmark classifier | EquiVAE on SVHN · SVHN test (supervised) | error rate: 10.04 ± 0.14 -> 11.97 ± 0.34 | Hurt |
| Use the invariant embedding distance classifier instead of the dedicated benchmark classifier | EquiVAE on MNIST · MNIST test (supervised) | error rate: 0.84 ± 0.03 -> 0.82 ± 0.05 | Improved |
| Use the invariant embedding distance classifier instead of the dedicated benchmark classifier | EquiVAE on SVHN · SVHN test (supervised) | error rate: 10.04 ± 0.14 -> 12.30 ± 0.28 | Hurt |
| Train the classifier jointly with the semi-supervised EquiVAE objective rather than on labelled data alone | EquiVAE · MNIST, 100 labels | error rate: 21.91± 0.66 -> 8.90± 0.70 | Improved |
| Train the classifier jointly with the semi-supervised EquiVAE objective rather than on labelled data alone | EquiVAE · MNIST, 600 labels | error rate: 6.64± 0.35 -> 3.99± 0.17 | Improved |
| Train the classifier jointly with the semi-supervised EquiVAE objective rather than on labelled data alone | EquiVAE · MNIST, 1000 labels | error rate: 5.43± 0.31 -> 3.34± 0.17 | Improved |
| Train the classifier jointly with the semi-supervised EquiVAE objective rather than on labelled data alone | EquiVAE · MNIST, 3000 labels | error rate: 2.96± 0.11 -> 2.23± 0.14 | Improved |
| Train the classifier jointly with the semi-supervised EquiVAE objective rather than on labelled data alone | EquiVAE · SVHN, 1000 labels | error rate: 39.64 ± 1.47 -> 37.95 ± 0.66 | Improved |
| Train the classifier jointly with the semi-supervised EquiVAE objective rather than on labelled data alone | EquiVAE · SVHN, 3000 labels | error rate: 25.50 ± 0.91 -> 24.95 ± 0.57 | Improved |
| Allow the invariant representation to receive the target x directly instead of using complementary same-class samples (ablation) | EquiVAE generative model · MNIST and SVHN (unsupervised training) | reconstruction of within-class structure | Hurt |
Closest papers by meaning
- Interpreting Equivariant Representations (ICML 2024 · similarity 0.90)
- Neural Fourier Transform: A General Approach to Equivariant Representation Learning (ICLR 2024 · similarity 0.89)
- Improving Equivariant Networks with Probabilistic Symmetry Breaking (ICLR 2025 · similarity 0.89)
- On Making $SE(2)$-Invariant Networks Optimal (NeurIPS 2026 · similarity 0.88)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.88)
- Approximately Equivariant Neural Processes (NeurIPS 2024 · similarity 0.88)
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.
