Research map / Learning theory and algorithms / Kernels and regression theory
Kernel Method for Persistence Diagrams via Kernel Embedding and Weight Factor
Genki Kusano, Kenji Fukumizu, Yasuaki Hiraoka
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 arctan persistence weight w_arc instead of w_pers (ablation) | PWGK (Gaussian kernel) · Synthetic XOR persistence-diagram classification | classification rate (%): 52.3 -> 85.6 | Improved |
| use arctan weight w_arc instead of unweighted w_one (ablation) | PWGK (Gaussian kernel) · Synthetic XOR persistence-diagram classification | classification rate (%): 55.1 -> 85.6 | Improved |
| use Gaussian kernel k_G instead of linear k_L as base kernel in PWGK (ablation) | PWGK w_arc p=5 · Synthetic XOR persistence-diagram classification | classification rate (%): 49.3 -> 75.8 | Improved |
| use Gaussian kernel on RKHS instead of linear kernel on PWGK embedding | PWGK k_G w_arc p=5 · Synthetic XOR persistence-diagram classification | classification rate (%): 75.8 -> 85.6 | Improved |
| vary degree p in w_arc among 1, 5, 10 (ablation) | PWGK · Synthetic XOR persistence-diagram classification | classification rate (%): 85.8 / 85.6 / 86.0 | No clear effect |
| increase persistence image mesh size from 20 to 100 with w_arc (ablation) | Persistence image with w_arc (Gaussian) · Synthetic XOR persistence-diagram classification | classification rate (%): 52.0 -> 85.8 | Improved |
| use PWGK instead of PSSK, persistence landscape, persistence image | PWGK · Synthetic XOR persistence-diagram classification | classification rate (%): PSSK 50%, PL 50%, PI 55% -> 85% | Improved |
| use PWGK instead of MTF on protein-drug binding | PWGK with SVM · Protein-drug binding (M2 channel) | CV classification rate (%): 98.31 -> 100 | Improved |
| use PWGK instead of MTF on hemoglobin | PWGK with SVM · Hemoglobin taut vs relaxed | CV classification rate (%): 84.50 -> 88.90 | Improved |
| use PWGK versus PSSK/persistence image for SiO2 glass transition KPCA | PWGK · SiO2 glass transition | cluster separation in KPCA (qualitative): change points l=39 (PWGK) vs l=33 (PSSK, PI) | Improved |
| apply all four methods to granular packing change point | PWGK, PSSK, landscape, image · Granular packing | KFDR change point: all detect l=23 | No clear effect |
Closest papers by meaning
- Persistence weighted Gaussian kernel for topological data analysis (ICML 2016 · similarity 0.94)
- Persistence Fisher Kernel: A Riemannian Manifold Kernel for Persistence Diagrams (NeurIPS 2018 · similarity 0.91)
- Towards a Persistence Diagram that is Robust to Noise and Varied Densities (ICML 2023 · similarity 0.91)
- Sliced Wasserstein Kernel for Persistence Diagrams (ICML 2017 · similarity 0.90)
- A Class of Topological Pseudodistances for Fast Comparison of Persistence Diagrams (AAAI 2024 · similarity 0.88)
- Wasserstein convergence of Cech persistence diagrams for samplings of submanifolds (NeurIPS 2024 · similarity 0.85)
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.
