Research map / Label-efficient and robust learning / Self-supervised and contrastive
Supervised Dimensionality Reduction and Visualization using Centroid-Encoder
Tomojit Ghosh, Michael Kirby
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 |
|---|---|---|---|
| add pre-training to centroid-encoder (ablation) | Centroid-encoder · MNIST | k-NN (k=5) error %: 3.17 -> 2.61 | Improved |
| add pre-training to centroid-encoder (ablation) | Centroid-encoder · USPS | k-NN (k=5) error %: 2.98 -> 2.91 | No clear effect |
| use centroid-encoder instead of autoencoder | autoencoder · MNIST | k-NN (k=5) error %: 22.04 -> 2.61 | Improved |
| use centroid-encoder instead of supervised UMAP | supervised UMAP · MNIST | k-NN (k=5) error %: 3.45 -> 2.61 | Improved |
| use centroid-encoder instead of nonlinear NCA | NNCA · USPS | k-NN (k=5) error %: 6.58 -> 2.91 | Improved |
| use centroid-encoder instead of dt-MCML (with pre-training) | dt-MCML · MNIST | k-NN (k=5) error %: 2.03 -> 2.61 | Hurt |
| use centroid-encoder instead of dt-MCML (with pre-training) | dt-MCML · USPS | k-NN (k=5) error %: 2.46 -> 2.91 | Hurt |
| use centroid-encoder instead of dt-MCML (no pre-training) | dt-MCML · USPS | k-NN (k=5) error %: 4.07 -> 2.98 | Improved |
| use centroid-encoder instead of HOPE (no pre-training) | HOPE · MNIST | k-NN (k=5) error %: 3.20 -> 3.17 | No clear effect |
| use centroid-encoder instead of SUMAP | SUMAP · Landsat | k-NN (k=5) error %: 9.68 -> 8.97 | No clear effect |
| use centroid-encoder instead of SNeRV lambda=0.1 | SNeRV · Letter | k-NN (k=5) error %: 22.96 -> 23.82 | Hurt |
| use centroid-encoder instead of S-Isomap | S-Isomap · Phoneme | k-NN (k=5) error %: 8.61 -> 7.52 | Improved |
Closest papers by meaning
- Hierarchical Nearest Neighbor Graph Embedding for Efficient Dimensionality Reduction (CVPR 2022 · similarity 0.82)
- Parameterized Low-Rank Regularizer for High-dimensional Visual Data (IJCV 2025 · similarity 0.80)
- The tree autoencoder model, with application to hierarchical data visualization (NeurIPS 2024 · similarity 0.79)
- Partition First, Embed Later: Laplacian-Based Feature Partitioning for Refined Embedding and Visualization of High-Dimensional Data (ICML 2025 · similarity 0.77)
- Beyond Worst-Case Dimensionality Reduction for Sparse Vectors (ICLR 2025 · similarity 0.77)
- Finsler Multi-Dimensional Scaling: Manifold Learning for Asymmetric Dimensionality Reduction and Embedding (CVPR 2025 · similarity 0.77)
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.
