Research map / Probabilistic and scientific modeling / Molecules and proteins
Persistence Images: A Stable Vector Representation of Persistent Homology
Henry Adams, Tegan Emerson, Michael Kirby, Rachel Neville, Chris Peterson, Patrick Shipman, Sofya Chepushtanova, Eric Hanson, Francis Motta, Lori Ziegelmeier
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 |
|---|---|---|---|
| concatenate H0 and H1 persistence images instead of H0 alone (ablation) | PI classifier, linked twist map · Linked twist map | classification accuracy: 49.8% -> 82.5% | Improved |
| concatenate H0 and H1 persistence images instead of H1 alone (ablation) | PI classifier, linked twist map · Linked twist map | classification accuracy: 65.7% -> 82.5% | Improved |
| use PIs instead of resized raw surfaces | subspace discriminant ensemble · aKS equation, t=3 | classification accuracy: 26.0% -> 72.7% | Improved |
| use PIs instead of surface height variance classifier | aKS classification · aKS equation, t=5 | classification accuracy: 75.2% -> 95.3% | Improved |
| add H0 to H1 persistence images (ablation) | aKS classification · aKS equation, t=10 | classification accuracy: 93.3% -> 97.3% | Improved |
| use PIs instead of PDs with L1 metric (H0) | K-medoids classification · Shape point clouds, noise 0.05 | computation time: 37346s -> 9s | Improved |
| use PIs instead of PDs with L-infinity metric (H1) | K-medoids classification · Shape point clouds, noise 0.05 | accuracy: 81.3% -> 100% | Improved |
| use PIs instead of PLs (H1, L2) | K-medoids classification · Shape point clouds, noise 0.1 | accuracy: 66.7% -> 96.0% | Improved |
| use PIs instead of PDs (H0, L1) | K-medoids classification · Shape point clouds, noise 0.05 | accuracy: 96.0% -> 93.3% | Hurt |
| vary PI resolution (5x5 to 100x100) and Gaussian variance (0.01 to 0.2) (ablation) | PI K-medoids · Shape point clouds | accuracy | No clear effect |
| vary Gaussian variance from 0.0001 to 0.1 (ablation) | H0 PIs, aKS · aKS equation | classification accuracy: less than one percentage point | No clear effect |
Closest papers by meaning
- The Persistent Laplacian for Data Science: Evaluating Higher-Order Persistent Spectral Representations of Data (ICML 2023 · similarity 0.91)
- Stable Vectorization of Multiparameter Persistent Homology using Signed Barcodes as Measures (NeurIPS 2023 · similarity 0.91)
- A Framework for Fast and Stable Representations of Multiparameter Persistent Homology Decompositions (NeurIPS 2023 · similarity 0.91)
- Persistent Homology through Image Segmentation (Student Abstract) (AAAI 2023 · similarity 0.90)
- Differentiability and Optimization of Multiparameter Persistent Homology (ICML 2024 · similarity 0.88)
- Graphcode: Learning from multiparameter persistent homology using graph neural networks (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.
