Skip to main content
. 2021 Jun 11;23(6):740. doi: 10.3390/e23060740

Figure 11.

Figure 11

Plot showing how the optimal value of the KD hyper-parameter K=σk·NS varies as a function of sample size NS and pdf type when using a Gaussian kernel. In disagreement with Parzen-window theory, the optimal value for K does not remain approximately constant as the sample size NS is varied. Further, the value of K varies significantly with shape of the underlying pdf.