Research map / Learning theory and algorithms / Matrix and tensor methods
Log-Euclidean Signatures for Intrinsic Distances Between Unaligned Datasets
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Compare task embeddings with LES rather than the strongest reported baseline, GW | MetaOptNet-SVM task-difficulty prediction · FC100 1-, 5-, and 10-shot | Pearson correlation with accuracy: -0.919±0.034 -> -0.934±0.034; -0.921±0.046 -> -0.945±0.032; -0.914±0.045 -> -0.935±0.032 | Helped |
| Compare task embeddings with LES rather than GW | MetaOptNet-SVM task-difficulty prediction · CIFAR-FS 1-shot | Pearson correlation with accuracy: -0.677±0.126 -> -0.671±0.128 | Mixed |
| Compare task embeddings with LES rather than GW | MetaOptNet-SVM task-difficulty prediction · CIFAR-FS 5- and 10-shot | Pearson correlation with accuracy: -0.672±0.167 -> -0.698±0.155; -0.582±0.171 -> -0.657±0.175 | Helped |
| Use LES instead of IMD to compare neural-network layer embeddings | All-CNN-C layer matching · CIFAR-10 identical inputs | Layer position classification accuracy: 85.2% -> 96% | Helped |
| Use LES rather than IMD when input correspondence is unavailable | All-CNN-C layer matching · CIFAR-10 different inputs | Layer position classification accuracy: 81% -> 95.8% | Helped |
| Vary the task-distance aggregation from mean to min or max (ablation) | LES task-difficulty prediction · FC100 and CIFAR-FS, 1-, 5-, and 10-shot | Pearson correlation with accuracy | No clear effect |
| Use LES instead of IMD to compare gene-space datasets with different gene dimensions | Single-cell developmental-stage comparison · scRNA-seq, different gene subsets across time points | Correlation with time: 0.27 -> 0.69 | Helped |
Closest papers by meaning
- Euclidean Embedding of Data Using Local Distances (NeurIPS 2026 · similarity 0.87)
- Recovering Manifold Structure Using Ollivier Ricci Curvature (ICLR 2025 · similarity 0.87)
- Wrapped Gaussian on the manifold of Symmetric Positive Definite Matrices (ICML 2025 · similarity 0.87)
- Diffusion Models and the Manifold Hypothesis: Log-Domain Smoothing is Geometry Adaptive (NeurIPS 2025 · similarity 0.86)
- Geometry-Aware Similarity Metrics for Neural Representations on Riemannian and Statistical Manifolds (NeurIPS 2026 · similarity 0.86)
- Diffusion Models Encode the Intrinsic Dimension of Data Manifolds (ICML 2024 · similarity 0.86)
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.