Scalable Affinity Propagation for Massive Datasets

Authors: Hiroaki Shiokawa9639-9646

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

Reproducibility Variable Result LLM Response
Research Type Experimental 4 Experimental Evaluation We experimentally evaluated the effectiveness of Scale AP by comparing it with the following AP algorithms. [...] Datasets: We used three real-world datasets summarized in Table 1. [...] Figure 1 shows the running time on the real-world datasets, where DNF indicates that the runtime exceeded 24 hours.
Researcher Affiliation Academia Hiroaki Shiokawa Center for Computational Sciences, University of Tsukuba, Japan shiokawa@cs.tsukuba.ac.jp
Pseudocode Yes Algorithm 1 Proposed method: Scale AP
Open Source Code No The paper does not provide any concrete statement or link regarding the availability of its source code.
Open Datasets Yes Datasets: We used three real-world datasets summarized in Table 1. All the datasets are published by UCI Machine Learning Repository (Dua and Graff 2017).
Dataset Splits No The paper mentions using real-world datasets but does not explicitly provide training, validation, or test dataset splits.
Hardware Specification Yes All experiments were conducted on a server with Intel Xeon CPU 2.60 GHz and 768 Gi B RAM.
Software Dependencies No The paper discusses various algorithms and methods but does not provide specific version numbers for software dependencies or libraries used in their implementation.
Experiment Setup Yes The experiments used the negative Euclidean distance as the similarity between data objects. In accordance with (Frey and Dueck 2007), we set the preference to both the median and minimum of the similarities, λ = 0.5, and the maximum number of iterations to T = 1, 000.