Neural Capacitated Clustering

Authors: Jonas K. Falkner, Lars Schmidt-Thieme

IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In our experiments on artificial data and two real world datasets our approach outperforms several state-of-the-art mathematical and heuristic solvers from the literature.
Researcher Affiliation Academia Jonas K. Falkner and Lars Schmidt-Thieme Institute of Computer Science, University of Hildesheim, Hildesheim, Germany {falkner, schmidt-thieme}@ismll.uni-hildesheim.de
Pseudocode Yes Algorithm 1: Capacitated k-means; Algorithm 2: Neural Capacitated Clustering (NCC)
Open Source Code Yes We open source our code at https://github.com/jokofa/NCC
Open Datasets Yes As real world datasets for capacitated spatial clustering we select the well-known Shanghai Telecom (ST) dataset [Wang et al., 2019]... The second dataset we assemble by matching the internet access sessions in the call record data of the Telecom Italia Milan (TIM) dataset [Barlacchi et al., 2015] with the Milan cell-tower grid retrieved from Open Celli D [OCID, 2021]... To evaluate our algorithm we choose the benchmark dataset of [Uchoa et al., 2017]
Dataset Splits No The paper mentions tuning parameters on a 'separate validation set' and discusses 'validation accuracy', but does not provide specific details on the size, percentage, or methodology of this validation split from the dataset.
Hardware Specification Yes All experiments are run on an i7-7700K CPU (4.20GHz).
Software Dependencies Yes We implement our model and the simple baselines in PyTorch [Paszke et al., 2019] version 1.11 and use Gurobi version 9.1.2 for all methods that require it.
Experiment Setup Yes We use L = 4 GNN layers, an embedding dimension of demb = 256 and a dimension of dh = 256 for all hidden layers... Then we do supervised training using binary cross entropy (BCE)... The fraction α of final nodes for which the re-prioritization is applied we treat as a hyperparameter... we set reasonable total run times of 3min for n = 200 and 15min for the original sizes.