Persistence weighted Gaussian kernel for topological data analysis

Authors: Genki Kusano, Yasuaki Hiraoka, Kenji Fukumizu

ICML 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the performance of the proposed kernel method with synthesized and real-world data, including protein datasets (taken by NMR and X-ray crystallography experiments) and oxide glasses (taken by molecular dynamics simulations). In Section 4, we apply KG for SVM, kernel PCA, and kernel change point detection.
Researcher Affiliation Academia Genki Kusano1 GENKSN@GMAIL.COM Kenji Fukumizu2 FUKUMIZU@ISM.AC.JP Yasuaki Hiraoka1 HIRAOKA@WPI-AIMR.TOHOKU.AC.JP 1Tohoku University, 2The Institute of Statistical Mathematics
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets No The paper mentions protein datasets (taken by NMR and X-ray crystallography experiments) and oxide glasses from molecular dynamics simulations, and refers to "data used in (Nakamura et al., 2015a;b)" for the Si O2 data. However, it does not provide concrete access information (links, DOIs, or explicit public availability statements) for these datasets in the paper itself.
Dataset Splits Yes The hyper-parameters (σ, C) in the PWGK and t in the PSSK are chosen by the 10-fold crossvalidation, and the degree p in the weight of the PWGK is set to be 5. SVMs are trained with persistence diagrams given by 100 data sets, and evaluated with 99 independent test data sets.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions that "all persistence diagrams are 1-dimensional (i.e., rings) and computed by CGAL (Da et al., 2015) and PHAT (Bauer et al., 2014)", but it does not specify version numbers for these software components.
Experiment Setup Yes In this paper, since all points of data are in R3, we set p = 5 from the assumption p > d + 1 in Theorem 3.2. For the parameter C, we also set C = (median{pers(Dℓ) | ℓ= 1, . . . , n}) p, where pers(D) = median{pers(xi) | xi D}. Similarly, τ is defined by median{dwarc k G (Di, Dj) | 1 i < j n}.