Interpretable Clustering via Multi-Polytope Machines
Authors: Connor Lawless, Jayant Kalagnanam, Lam M Nguyen, Dzung Phan, Chandra Reddy7309-7316
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We benchmark our approach on a suite of synthetic and real world clustering problems, where our algorithm outperforms state of the art interpretable and non-interpretable clustering algorithms. |
| Researcher Affiliation | Collaboration | 1 Cornell Univeristy, Operations Research and Information Engineering, Ithaca, NY 14850 2 IBM Research, Thomas J. Watson Research Center, Yorktown Heights, NY 10598, USA |
| Pseudocode | Yes | Algorithm 1: Cluster Initialization via Alternating Minimization Algorithm 2: Multi-Polytope Clustering (MPC) Algorithm |
| Open Source Code | No | The paper does not provide a specific link or explicit statement about the public release of its own source code for the methodology described. |
| Open Datasets | Yes | We tested our algorithm on two sets of clustering datasets. The first is a set of 9 synthetic clustering instances called the fundamental clustering and projection suite (Ultsch and L otsch 2020). ... a set of 8 datasets from the UCI machine learning repository. |
| Dataset Splits | No | The paper does not explicitly provide specific training, validation, and test dataset splits (e.g., percentages or sample counts). It mentions tuning 'k' and running algorithms multiple times but not data partitioning for validation. |
| Hardware Specification | No | The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions comparing against various algorithms (e.g., k-means++, GMM, DBSCAN, ICOT, Ex KMC) and notes that supplementary materials contain descriptions of implementations, but it does not specify software dependencies with version numbers for reproducibility of the authors' own work. |
| Experiment Setup | Yes | The first, MPC-1, sets M = β = 1 and represents cluster explanations with only axis-parallel hyperplanes... The second, MPC-2, sets M = 3, β = 2 which allows for more general hyperplanes with up to two non-zero integer coefficients and coefficients within [ 3, 3]. ... For all algorithms that require a specification of the number of clusters we tune k between 2 and 10. We ran each algorithm 100 times with different random seeds and report the best result. |