• This repository provides our implementation of D-DPC.
• This is a fast algorithm for density-peaks clustering (proposed in Science) on fully dynamic data.
• As for the details about this algorithm, please read our paper, Efficient Density-Peaks Clustering Algorithms on Static and Dynamic Data in Euclidean Space.

• spatial library
  • We used version 2.1.8.
  • Path to this library should be changed based on your environment, see ddpc.hpp.
• The source codes of D-DPC have to be changed based on your paths of the above library.
• g++ 7.4.0 (or higher version) and Openmp.

• We prepared codes for Ubuntu (20.04 LTS).
• We assume low-dimensional datasets, as we use a kd-tree.
  • Set dimensionality in file_input.hpp.
• Compile: g++ -O3 main.cpp -o ddpc.out -fopenmp.
• Run: ./ddpc.out.

• Our code inputs csv file, where each row shows a vector whose elements are separeted by ,.
  • For example, one 3-dimensional data is described by 1.0,10,2.5.
• For your data, assign dataset_id as written in input_data() function of data.hpp to read your dataset.
• The path of your dataset can be freely changed in data.hpp, see input_data() function.
• To simulate random point insertions and deletions, we sorted a given dataset by a random order. If you need to remove arbitrary points, please change our codes.

• Set some value in the corresponding txt file in parameter.

• Initialization is done by Ex-DPC++ like algorithm (the dependent point of each point is computed by incremental kNN search).
  • initialization() function is much similar to Ex-DPC++ (the main difference is that we incrementally compute the kNNs of each point online).

If you use our implementation, please cite the following paper.

@article{amagata2024dpc,  
	title={Efficient Density-Peaks Clustering Algorithms on Static and Dynamic Data in Euclidean Spaces},  
	author={Amagata, Daichi and Hara, Takahiro},  
	booktitle={ACM Transactions on Knowledge Discovery from Data},
	volume={18},
	number={1},
	pages={2:1--2:27},  
	year={2024}  
}

Copyright (c) 2021 Daichi Amagata
This software is released under the MIT license.