Research map / Label-efficient and robust learning / Recognition and long-tail
Persistence Paths and Signature Features in Topological Data Analysis
Ilya Chevyrev, Vidit Nanda, Harald Oberhauser
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 Betti embedding in feature form instead of sliced Wasserstein kernel | k_SW · Shapes | accuracy: 95.8 -> 98.1 | Improved |
| use Euler embedding feature map instead of persistence image features | Phi_PI · Shapes | accuracy: 90.3 -> 98.0 | Improved |
| use Betti kernel instead of sliced Wasserstein kernel | k_SW · Textures | accuracy: 96.8 -> 97.8 | Improved |
| use signature of persistence paths instead of persistence images on orbits | Phi_PI · Orbits | accuracy: 99.86 -> 98.8 | Hurt |
| kernelize Betti embedding instead of direct feature map (ablation) | Phi_beta · Textures | accuracy: 96.6 -> 97.8 | Improved |
| kernelize envelope embedding instead of direct feature map (ablation) | Phi_E · Textures | accuracy: 88.1 -> 90.4 | Improved |
| use direct feature map instead of kernel for Euler embedding (ablation) | k_chi · Shapes | accuracy: 92.4 -> 98.0 | Improved |
| use direct feature map instead of kernel for envelope embedding (ablation) | k_E · Shapes | accuracy: 92.7 -> 95.0 | Improved |
| use direct feature map instead of kernel for envelope embedding (ablation) | k_E · Orbits | accuracy: 96.6 -> 98.1 | Improved |
| use landscape or naive path embedding instead of envelope, Betti, Euler (ablation) | signature feature map · Textures, Orbits, Shapes | accuracy | Hurt |
| raise tensor truncation level M for feature maps (up to 8) vs kernels (2 or 3) (ablation) | signature features · Textures, Orbits, Shapes | best M | Mixed |
Closest papers by meaning
- Stable Vectorization of Multiparameter Persistent Homology using Signed Barcodes as Measures (NeurIPS 2023 · similarity 0.92)
- Learning Representations of Persistence Barcodes (JMLR 2019 · similarity 0.92)
- A Survey of Vectorization Methods in Topological Data Analysis (TPAMI 2023 · similarity 0.91)
- Differentiability and Optimization of Multiparameter Persistent Homology (ICML 2024 · similarity 0.91)
- Graphcode: Learning from multiparameter persistent homology using graph neural networks (NeurIPS 2024 · similarity 0.91)
- Persistent Homology for High-dimensional Data Based on Spectral Methods (NeurIPS 2024 · similarity 0.87)
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.
