Shared-Nearest-Neighbor-Based Clustering by Fast Search and Find of Density Peaks

C++ and Python implementation of SNNDPC algorithm.

The Matlab version is moved to the branch MatlabImplementation.

Demo

If you use Windows:

1. Open a Visual Studio developer command prompt
   • Because of the toolchain paths
2. cd to project root SNNDPC/
3. cmake -DCMAKE_BUILD_TYPE=Release -G "NMake Makefiles" -S . -B build\release
4. cmake --build build\release --target Demo
5. build\release\Demo.exe

The demo runs on the S2 dataset.

If CMake complains about its version, modify SNNDPC/CMakeLists.txt:1 to fit your CMake version.

To use other datasets, see Customization

Environment

Some highlighted requirements.

• Python: 3.8, because of := syntax
• IntelTBB: Optional, for parallelization
• OpenMP: Optional, for parallelization

Customization

Provided Dataset

To use other provided datasets in demo:

1. Modify variable pathData at SNNDPC/Demo.cpp:11.

   • Macro SOLUTION_DIR is the absolute path to SNNDPC/.

2. Modify variables k, n, d, and nc according to the paper

   | Variable | Reference | | -------- | --------------------------------- | | k | Table 4, column Arg- | | n | Table 2, column No. of records | | d | Table 2, column No. of attributes | | nc | Table 2, column No. of clusters |

External Dataset + Demo.cpp

To use external datasets in Demo.cpp:

1. Make sure your dataset has exactly 3 columns: x, y, and label.
   • If you want to use more attributes, you need to edit the fscanf call at Demo.cpp:24.

External Dataset + Custom Runner

To use external datasets in a custom runner:

1. Load your dataset into a (flattened) 1D C-style array data, shape [n×d].
   • For the logical 2D array, each row is a record, and each column is an attribute.
2. Include SNNDPC.hpp.
3. Call the SNNDPC() with parameters k, n, d, nc, and data.
4. The function will return two pointers, the centroids and the assignment.
   • Both are a 1D array.
   • Both are created by new, you need to manually delete them to prevent memory leak.

Contact

If you have any inquiries, please open an issue instead of sending emails directly to me. My email address on the paper is no longer frequently checked.