Fig. 4 | Scientific Reports

Fig. 4

From: A meta-learning framework to mitigate negative transfer in transfer learning applicable to drug design

Fig. 4

Model performance for the first calculation setting. (A) Shown are AUC boxplots (box: 1 st quartile, median, 3rd quartile; whiskers: +/− 1.0 x interquartile range; dots: outliers) for 50 independent trials of meta- and standard transfer learning models for the 19 target PKs. Statistical significance is indicated by asterisks; \(\:0.05<\text{p-value}\le\:1\): ns (no statistical significance), \(\:0.01<\text{p-value}\le\:0.05\): *, \(\:0.001<\text{p-value}\le\:0.01\): **, and \(\:\text{p-value}\le\:0.001\): ***. (B) Boxplots representing the distribution of differences in AUC (dAUC) on a per-trial basis are shown. Positive values indicate improved performance of meta-learning compared to standard transfer learning models.

Back to article page