where \(F^r\), \(S^r\) represent the task similarity sequence from the property relation map and transferability sequence of the \(r\)-th task, \(F^r[t]\), \(S^r[t]\) indicate the relation rank number of the \(t\)-th task for the \(r\)-th task, \(T\) represents the number of source tasks (actually \(T=41\)). The scope of the above cosine similarity is from 0 to 1, and a high value means the similarity sequence \(F^r\) greatly approximates \(S^r\).
Finally, we calculated the cosine similarity result for each target task and then averaged these results with equation (3) as follows: