Dynamic Facility Location in High Dimensional Euclidean Spaces

Authors: Sayan Bhattacharya, Gramoz Goranci, Shaofeng H.-C. Jiang, Yi Qian, Yubo Zhang

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on real datasets confirm that our algorithm achieves high-quality solutions with low running time, and incurs minimal recourse.
Researcher Affiliation Academia 1University of Warwick, UK 2Faculty of Computer Science, University of Vienna, Austria 3Peking University, China.
Pseudocode Yes Algorithm 1 (Czumaj et al., 2024) on input point-set P... Algorithm 2 INSERT(p)
Open Source Code No The paper mentions 'Our implementation' but does not state it is open-source or provide a link to the code for the described methodology.
Open Datasets Yes Our experiment is done on Twitter (Chan et al., 2018b), Census1990 (Meek et al.), Covertype (Blackard, 1998) and KDD-Cup (Stolfo et al., 1999) datasets.
Dataset Splits No The paper specifies using a 'sliding window of size ℓ= 1000' but does not provide explicit training, validation, or test dataset splits (e.g., percentages or counts).
Hardware Specification Yes All the experiments are conducted on an Apple computer with M1 Pro CPU and 16GB memory.
Software Dependencies No The paper mentions implementation 'in C++ language' and use of 'standard locality-sensitive hashing (LSH) techniques' but does not specify version numbers for any software or libraries.
Experiment Setup Yes We consider a sliding window of size ℓ= 1000, and our update operations on the datasets are generated by this sliding window. The opening cost is set such that a moderate number of facilities would be open in a (near-)optimal solution. ... In our experiments, we use 15 random hash functions and we find this setup already produces decent accuracy when combining with our algorithm.