Research on Kilosort4 spike clustering algorithm based on mini batch K-means optimization
ZHOU Fuhao
LI Zhaochun
WANG Yucheng
Abstract：To address the time complexity bottleneck caused by the traditional K-means algorithm in the template deconvolution stage,when Kilosort4 processes spike data.We proposed an optimization method based on mini batch K-means clustering algorithm.Firstly,the K-means++algorithm was used to initialize cluster centers.Then,during the iterative process,a dynamic data subset was extracted,local clustering was performed and the size of the subset was adaptively adjusted according to the size of the data set.Final-ly,the cluster centers were updated incrementally until convergence.Experimental results demonstrated that the optimized Kilosort4 al-gorithm achieved approximately 8%improvement in processing speed compared to the original algorithm.This algorithm can significant-ly reduce computational complexity while maintaining clustering quality.This research can provide more efficient tools for neuroscience studies.
Keywords：SpikeMini batch K-means algorithmKilosort4Speed optimizationNeuroscienceElectrophysiological signal
Publication Date：2026-04-30
Online Publishing Date：2026-09-11(First online date of this platform, not the publication date of the document)
Pages：6( 98-103 )