Hypergraph p-Laplacian: A Differential Geometry View

Authors: Shota Saito, Danilo Mandic, Hideyuki Suzuki

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed p-Laplacian is shown to outperform standard hypergraph Laplacians in the experiment on a hypergraph semisupervised learning and normalized cut setting. Our experiment on hypergraph semi-supervised clustering problem shows that our hypergraph p-Laplacian outperforms the current hypergraph Laplacians.
Researcher Affiliation Academia Shota Saito The University of Tokyo ssaito@sat.t.u-tokyo.ac.jp Danilo P Mandic Imperial College London d.mandic@imperial.ac.uk Hideyuki Suzuki Osaka University hideyuki@ist.osaka.ac.jp
Pseudocode No The paper describes algorithms and update rules using mathematical equations and textual descriptions, but it does not provide structured pseudocode blocks or algorithm figures.
Open Source Code No The paper does not provide any statement or link regarding the availability of its source code.
Open Datasets Yes We summarize the benchmark datasets we used in Table 1. All datasets were taken from UCI Machine Learning Repository.
Dataset Splits Yes The parameter μ was chosen for all methods from 10k, where k {0, 1, 2, 3, 4} by 5-fold cross validation. We randomly picked up a certain number of labels as known labels, and predicted the remaining ones. We repeated this procedure 10 times for different number of known labels.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as CPU or GPU models.
Software Dependencies No The paper does not specify any software dependencies with version numbers.
Experiment Setup Yes The parameter μ was chosen for all methods from 10k, where k {0, 1, 2, 3, 4} by 5-fold cross validation. For our p-Laplacian, we varied p from 1 to 3 with the interval of 0.1, and we show the result of p=2 and the result of p giving the smallest average error for each number of known labeled points. The parameter p for Hein s regularizer is fixed at 2, since this is recommended by Hein et al. (2013). We take ψ(0) = y as an initial condition.