Skip to main content
. 2020 May 19;6(6):939–949. doi: 10.1021/acscentsci.0c00229

Figure 2.

Figure 2

Effect of training set sample size on model generalizability. (a) Mean values for test set recalls computed using different sample sizes. Values approach 0.90 for all targets, when the training set size is within 250 000 and 1 million molecules. (b) Variations of standard deviations (STD) approach 0, for a sample size of 1 million molecules. We ran one iteration for each target and repeated computations five times at each sampling size.