Approximate Decomposable Submodular Function Minimization for Cardinality-Based Components
Authors: Nate Veldt, Austin R. Benson, Jon Kleinberg
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our sparse reduction technique leads to significant improvements in theoretical runtimes, as well as substantial practical gains in problems ranging from benchmark image segmentation tasks to hypergraph clustering problems." and "5 Experiments |
| Researcher Affiliation | Academia | Nate Veldt nveldt@tamu.edu Texas A&M University Austin R. Benson arb@cs.cornell.edu Cornell University Jon Kleinberg kleinberg@cornell.edu Cornell University |
| Pseudocode | Yes | Algorithm 1 GREEDYPLCOVER(g, ε)" and "Algorithm 2 SPARSECARD(f, ε) |
| Open Source Code | Yes | Code for our algorithms and experimental results is available at https://github.com/nveldt/Sparse Card DSFM. |
| Open Datasets | Yes | We consider the smallplant and octopus segmentation tasks from Jegelka et al. [20, 19]." and "Image datasets: http://people.csail.mit.edu/stefje/code.html. Hypergraph clustering datasets: www.cs.cornell.edu/~arb/data/. |
| Dataset Splits | No | The paper focuses on solving a minimization problem on given datasets (image segmentation, hypergraph clustering instances) rather than training a supervised model. It does not provide explicit train/validation/test splits for these datasets. |
| Hardware Specification | Yes | We ran experiments on a laptop with a 2.2 GHz Intel Core i7 processor and 8GB of RAM. |
| Software Dependencies | No | The paper mentions comparing against 'recent C++ implementations' and links to other GitHub repositories but does not list specific version numbers for compilers, libraries, or other software dependencies. |
| Experiment Setup | Yes | SPARSECARD was run for a range of ε values on a decreasing logarithmic scale from 1 to 10 4..." and "ACDM depends on a hyperparameter c controlling the number of iterations in an outer loop." and "We seek to detect 45 labeled clusters using a random seed set of 5% of each cluster. |