Towards faster seeding for k-means++ via lower bound and triangle inequality

This repository is the official implementation of [Towards faster seeding for k-means++ via lower bound and triangle inequality], submitted to Neurocomputing （Under Review））

Usage

The specific algorithm is implemented in the kmeans_plus_plus.py file.

To evaluate the efficiency of the proposed method, run 'main.py'

The Parameter Sensitivity Analysis can be verified by changing the parameters used in 'parameter_sensivity.py'. The MixedSinx file is too large. If you need it, you can email me to request it.

The interface designed for the UCR time series.

download and read the UCR dataset(https://www.cs.ucr.edu/~eamonn/time_series_data_2018/)

In the main file, just set the file_name to the time series file name you want to test.

In addition to the datasets in the paper, we have conducted experiments on all datasets in the UCR collection with more than 700 samples (53 newly added datasets). The experimental results, which include the runtime of each algorithm, can be found in the "results" folder. We have also generated critical difference diagrams for the newly added datasets, which can similarly be found in the "results" folder.

General Example

Two datasets 'coil100'， 'Wafer' are provided. Just run 'main.py'

Requirements: NumPy, scipy, matplotlib, math, cv2, pandas,PIL