Skip to main content
Figure 6 | Journal of Cheminformatics

Figure 6

From: Inferring multi-target QSAR models with taxonomy-based multi-task learning

Figure 6

Performance varying the supplied similarities. Mean squared error (MSE) while varying the supplied similarities. Each boxplot visualizes the performance on the 10 randomly generated test splits. True stands for correct task similarities given by k c o s , Anti for anti correlated similarities given by 1−k c o s , and Random for random similarities. The upper graph shows results for a high task similarity between the simulated tasks, the lower graph for a low similarity.

Back to article page