Fast topological clustering with Wasserstein distance

Authors: Tananun Songdechakraiwut, Bryan M Krause, Matthew I Banks, Kirill V Nourski, Barry D Van Veen

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed method is demonstrated to be effective using both simulated networks and measured functional brain networks.
Researcher Affiliation Academia Tananun Songdechakraiwut Department of Electrical and Computer Engineering University of Wisconsin Madison, USA songdechakra@wisc.edu Bryan M. Krause & Matthew I. Banks Department of Anesthesiology Department of Neuroscience University of Wisconsin Madison, USA Kirill V. Nourski Department of Neurosurgery Iowa Neuroscience Institute University of Iowa, USA Barry D. Van Veen Department of Electrical and Computer Engineering University of Wisconsin Madison, USA
Pseudocode No The paper describes an iterative algorithm in Section 3.2 but does not provide it in a formally structured pseudocode or algorithm block.
Open Source Code Yes Code for topological clustering is available at https://github.com/topolearn.
Open Datasets Yes We evaluate our method using an extended brain network dataset from the anesthesia study reported by Banks et al. (2020)
Dataset Splits No This paper focuses on clustering, not supervised learning, and thus does not describe traditional train/validation/test splits. It evaluates clustering performance against ground truth labels for the entire dataset used in the experiments.
Hardware Specification No The paper does not specify any hardware details such as GPU models, CPU types, or memory used for running the experiments.
Software Dependencies No The paper lists several external tools and their sources (e.g., GUDHI, B-ADMM, bctnet, GraKeL) that were used for baseline comparisons. However, it does not provide specific version numbers for the core software dependencies or environment (e.g., Python, TensorFlow/PyTorch, NumPy versions) used for their own proposed method.
Experiment Setup Yes Initial clusters for all methods are selected at random. ... We use µ = 1 and σ = 0.5 universally throughout the study. ... We calculate these performance metrics by running the algorithm for 100 different initial conditions