Hierarchical Multiple Kernel Clustering

Authors: Jiyuan Liu, Xinwang Liu, Siwei Wang, Sihang Zhou, Yuexiang Yang8671-8679

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments are conducted on benchmark datasets. It can observed that the proposed algorithm outperforms other representative MKC algorithms.
Researcher Affiliation Academia Jiyuan Liu, 1 Xinwang Liu, 1* Siwei Wang, 1 Sihang Zhou, 2 Yuexiang Yang 1 1 College of Computer, National University of Defense Technology, Changsha, Hunan, China 2 College of Intelligence Science and Technology, National University of Defense Technology, Changsha, Hunan, China {liujiyuan13, xinwangliu, wangsiwei13}@nudt.edu.cn, sihangjoe@gmail.com, yyx@nudt.edu.cn
Pseudocode Yes Algorithm 1: Hierarchical multiple kernel clustering
Open Source Code Yes Meanwhile, we open the source code of HMKC on Github9. 9https://github.com/liujiyuan13/HMKC-code release
Open Datasets Yes We evaluate the proposed algorithm on eight widely used MKC benchmark datasets, including AR10P1, BBCSport2, CCV3, Flower174, Flower1025, Ionosphere6, Heart7 and Plant8. Their details are shown in Table 1.
Dataset Splits No The paper does not explicitly describe a validation dataset split (e.g., 'validation set' or 'validation split') or cross-validation methodology for model training or selection. The term 'validation' is used in the context of 'Validation on Intermediary Matrix' to discuss the approach itself, not a data split for model evaluation.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., Python, PyTorch, scikit-learn versions) required to replicate the experiments.
Experiment Setup Yes Only the sizes of the intermediary matrices, {H(t) p }s t=1 are supposed to be given manually. In the following, we consider two experimental settings. The first is called HMKC-1 with employing 1-layer of intermediary matrices, {H(1) p }m p=1 Rn c, where c is searched from [2k, 3k, , 20k]. While the second one is named HMKC-2 with employing 2-layer of intermediary matrices, {H(1) p }m p=1 Rn c1 and {H(2) p }m p=1 Rn c2, in which c1 c2 and c1, c2 are grid searched from [2k, 3k, , 20k]. In specific, the search region is restricted to [2k, 3k, , 10k] for AR10P since it has relatively small number of samples in each cluster.