Skip to main content
. 2019 Nov 20;10:1145. doi: 10.3389/fgene.2019.01145

Figure 4.

Figure 4

Three models of Deepprune are compared on the simulated datasets. We show the performance of Deepprune based on different criteria with 4, 8, and 16 kernels. The performance of three different criteria is almost identical to that with 8 and 16 kernels during iterative pruning. However, the final model with four kernels shows that weighted-based Deepprune is superior to the other two methods in simulated dataset 1, but hard to classify owing to high entropy.