Figure 3From: Inferring multi-target QSAR models with taxonomy-based multi-task learningGraph-regularized multi-task (GRMT) SVR training. The example shows four tasks, represented by four nodes of a graph, and their corresponding weight vectors w 1 ,…, w 4 . The tasks are related by a real-valued adjacency matrix A. GRMT trains the task specific models w 1 ,…, w 4 in a single step, indicated by a dashed box, using the instances of all tasks.Back to article page