Quadratic Decomposable Submodular Function Minimization
Authors: Pan Li, Niao He, Olgica Milenkovic
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments in semi-supervised learning on hypergraphs confirm the efficiency of the proposed algorithm and demonstrate the significant improvements in prediction accuracy with respect to state-of-the-art methods. |
| Researcher Affiliation | Academia | Pan Li UIUC panli2@illinois.edu Niao He UIUC niaohe@illinois.edu Olgica Milenkovic UIUC milenkov@illinois.edu |
| Pseudocode | Yes | Algorithm 1: RCD Solver for (6) and Algorithm 2: The Conic MNP Method for Solving (9) |
| Open Source Code | Yes | The code for QDSFM is available at https://github.com/lipan00123/QDSDM. |
| Open Datasets | Yes | We also evaluated the proposed algorithms on three UCI datasets: Mushroom, Covertype45, Covertype67, used as standard datasets for SSL on hypergraphs [33, 17, 18]. |
| Dataset Splits | No | While the paper describes how labeled data is used ("uniformly at random picked l = 1, 2, 3, 4 vertices from each cluster to represent labeled datapoints" and "set the number of observed labels to 100"), it does not explicitly specify comprehensive training, validation, and test dataset splits by percentage or absolute counts for reproducibility of data partitioning. |
| Hardware Specification | Yes | The CPU times of all methods are recorded on a 3.2GHz Intel Core i5. |
| Software Dependencies | No | The paper states that methods are "implemented via C++" and "Inv Lap is executed via matrix inversion operations in Matlab", but it does not provide specific version numbers for these software components or any associated libraries. |
| Experiment Setup | Yes | For synthetic data: 'We uniformly at random generated 500 hyperedges within each cluster and 1000 hyperedges across the two clusters. Each hyperedge includes 20 vertices. We also uniformly at random picked l = 1, 2, 3, 4 vertices from each cluster to represent labeled datapoints.' and 'tuned β to be nearly optimal for different objectives with respect to the classification error rates. In the experiment, we used Wii = Dii, i'. For real data: 'We set β = 100 and Wii = 1, for all i, and set the number of observed labels to 100'. |