Fast kernel <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si10.svg"><mml:mi>k</mml:mi></mml:math>-means clustering using incomplete Cholesky factorization

Kernel-based clustering algorithms can identify and capture the nonlinear structure in datasets, thereby achieving better performance than linear clustering. However, large amount of memory involved computing storing entire kernel matrix makes it difficult for kernel-based to handle large-scale datasets. In this study, we prove that an incomplete Cholesky factorization (InCF) generate approximate with rank s a after iterations. We also show approximation error decreases exponentially as number iterations increases when exponential decay eigenvalues is sufficiently fast. therefore employ InCF accelerate save space. The key idea proposed k-means using approximated product low-rank its transpose. Then, applied columns transpose matrix. achieved by method increases. Experimental results algorithm similar algorithm, but be